Closed Bug 28237 Opened 26 years ago Closed 25 years ago

remove non-compliant text/xul mimetype.

Categories

(Core :: DOM: HTML Parser, defect, P3)

defect

Tracking

()

RESOLVED FIXED
mozilla1.0

People

(Reporter: braden, Assigned: braden)

References

Details

Attachments

(7 files)

The content type currently used to identify XUL ("text/xul") is not a valid type (that is, no such type has been assigned by IANA/ICANN) or an experimental type.
Status: NEW → RESOLVED
Closed: 26 years ago
Resolution: --- → REMIND
Yeah-- fine. But it's an internal type, and I'm not altogether sure that it matters.
What do you mean "it's an internal type"? XUL is a new XML application that Mozilla will be presenting to Web developers everywhere. What's internal about that? Mozilla is Open Source. It should follow standard protocols, not trample them. This matters.
Status: RESOLVED → REOPENED
Resolution: REMIND → ---
I'll take this.
Status: REOPENED → ASSIGNED
Target Milestone: M19
XUL is currently an internal XML schema that NETSCAPE invented for it's own purpose and (naturally) contributed to mozilla. We're not presenting this to web developers at this time; nor are we preventing them from using it. However, a lot of work must occur before this becomes anything like a web standard. Hyatt: What is your intention?
Then why are you impinging upon the MIME namespace by adopting their notation and usage for something you seem to claim is unrelated? In fact, it *isn't* unrelated, and RFC2048 makes ample accommodations for this situation. The type "text/xul" is wrong and should be corrected. Your comments confirm that it belongs in the vendor subtree, though they call into question whether the vendor would be mozilla.org or Netscape. I should add here that the notion that XUL falls under the authority of Netscape and not mozilla.org is disturbing--I was not aware that Mozilla would be supporting closed, proprietary formats. Finally, your assertion that XUL has not been presented to Web developers is patently wrong--it has been presented widely both through the open source process and documentation among the mozilla.org Web pages. The notion that a format must be a "Web standard" (quoted because I'm not really sure what the qualifications for that title are) to have a registered MIME type is also probably fallacy--just look at the list of existing types: <URL:http://www.isi.edu/in-notes/iana/assignments/media-types/media-types>
I agree with your assertion that using a vendor specific mimetype is the answer. As to your point about XUL/web developers: you're grossly underestimating the necessity of working within the realm of standards. It may be some time before we can work through the W3C processes to establish XUL as a standard. Netscape's history is littered with proprietary extensions, and it is not our objective to further fractionalize the web by unilaterally mandating new ones. Netscape plans to utilize XUL for it's own purposes, and invites third party developers to do the same. We will officially support XUL as a standard if and when it becomes one. Mozilla is free to act as it sees fit -- and you are of course free to behave according to your conscience.
I think we just might be in agreement. I have made no assertion that XUL should be presented to the W3C for standardization. Erik van der Poel is the one who presented this *possibility* on the newsgroup (<URL:news://news.mozilla.org/38A9C226.7DF332C0%40netscape.com>). My contentions are: * The "xul" subtype belongs in the IETF subtree if and only if XUL is a vendor-independent standard. (And I do consider mozilla.org a vendor and not a standards body.) * As long as XUL is not a vendor-independent standard, the "xul" subtype belongs in the vendor subtree when it is registered. * In either case, a MIME type needs to be registered--unless a subtype in the experimental subtree is used (and I do not think that is a desirable alternative). Thus, my agreement with Erik's proposal of "text/vnd.moz.xul" (<URL:news://news.mozilla.org/38A9C89E.89DABF6A%40netscape.com>).
My intent was to ignore it for a while (probably forever). I don't think this is a big deal at all. I think we could call it "foopy/goopy" and it wouldn't make a damn bit of difference.
That's nice. What, pray tell, is the grounds for the obvious aversion to registering a MIME type?
I'm telling you I don't give a rat's ass. I have more important things to worry about than the MIME type of XUL, like, say, making things actually work. Call it whatever you want.
Yes, it's quite apparent that there's a good deal of nonchalance within Netscape where MIME type procedures are concerned. A legacy of the historical disregard for Web standards Rick spoke of, I suppose. Whether failure to RTFM or deliberate disregard of established convention are the source of the current bogus type, I can only guess. I just find it peculiar that I'm the one here accused of "grossly underestimating the necessity of working within the realm of standards".
In this case, it was failure to RTFM.
Handing over to hyatt to address. Thanks David.
Assignee: rickg → hyatt
Status: ASSIGNED → NEW
Looking for a patch... also pulling in slightly...
Status: NEW → ASSIGNED
Whiteboard: [HELP WANTED]
Target Milestone: M19 → M18
It's nice to hear that most people have reached a consensus that "text/vnd.moz.xul" should be used. FWIW, Braden, I wholehartedly agree with you here. You're not the only one who cares about this. <rant>People who disagree: you have to realize that saying things like "it's an internal type" just isn't appropriate in an open source project. There's no such thing as "internal" here. It doesn't matter if Netscape invented it. At some point, the technology was made part of MOZILLA, and at that moment, it became very *public*. It's not an endorsed standard, but that doesn't give license to ignore the RFCs. It's still public, so it should follow the rules.</rant> Also, to get an idea of how things are from a technical standpoint, I did a little check: cx824225-a:~/src/mozilla$ fgrep -rl "text/xul" * docshell/base/nsDSURIContentListener.cpp gfx/src/nsImageNetContextAsync.cpp htmlparser/src/nsIParser.h layout/build/nsLayoutDLF.cpp mailnews/base/src/nsMessenger.cpp mailnews/base/src/nsMsgWindow.cpp mailnews/compose/src/nsMsgQuote.cpp mailnews/mime/build/nsMsgMimeCID.h mailnews/mime/emitters/build/nsMimeEmitterCID.h mailnews/mime/src/nsStreamConverter.cpp mailnews/mime/tests/mimetest/mimetest.cpp netwerk/mime/public/nsMimeTypes.h rdf/content/src/nsXULDocument.cpp xpfe/browser/src/nsBrowserInstance.cpp xpfe/components/directory/nsDirectoryViewer.cpp cx824225-a:~/src/mozilla$ That's 15 files -- not *too* bad. Maybe I'll make a patch. I'm not sure it's as simple as replacing "text/xul" everywhere, tho. I'll have to investigate.
If you make a patch and it passes the pre-checkin tests, I will accept it.
Thanks for the assistance here, Keith. Note that this bug report speaks to the validity of the type in use. Changing to another unregistered type--unless the change is to an experimental type--shouldn't affect the status of this bug report.
Right. The bug can be marked fixed as soon as the type is registered, which will probably be after the change is checked in. Who should do that? Me? Braden? Dave? Anybody? I've got the changes sitting here, but they're for vnd.moz.xul, and I've got to update that to vnd.mozilla.xul. Once that's done, (and I rebuild and test), I'll post the patch.
Well, there it is. It passes the pre-checkin tests in my tree. Once it is checked in, someone should make an announcement on the newsgroups to let everybody know to reconfigure their web servers. Now we just need to get it registered.
Could someone summarize why we need a different content type for cached XUL? Are we really talking about different representations of the data?
Folks, this document has just come to my attention: <URL:http://www.ietf.org/internet-drafts/draft-murata-xml-02.txt> In light of section 6 in this document, I recommend "text/vnd.mozilla.xul-xml".
Also of potential importance, from section 3 of the aforementioned document: Text/xml and application/xml behave differently when the charset parameter is not explicitly specified. If the default charset (i.e., US-ASCII) for text/xml is inconvenient for some reason (e.g., bad WWW servers), application/xml provides an alternative (see "Optional parameters" of "3.2 Application/xml Registration"). The same rules apply to the distinction between text/xml-external-parsed-entity and application/xml-external-parsed-entity. Unless XUL has any need for UTF-16, UCS-4, or UTF-32 (which I am inclined to doubt), this would appear to be the critical factor in the decision to use "text" or "application" as the primary type. Could some XUL authority (Dave?) comment on XUL's requirements in this regard?
Okay, if we want to use the "-xml" suffix, there's a patch for it. If you're applying this patch, be sure to do this one *instead* of the earlier patch, *not* in addition to it. Also note that I left the cached xul type as "text/vnd.mozilla.cached-xul" since it should not be touched by general-purpose xml parsers. (It may not even be xml data if it's pre-parsed or something. I don't know.) Also, don't worry about the cached xul type. It's only used by layout/xptoolkit to communicate within mozilla. The parser and necko don't even know anything about it, so if a web server sends "text/vnd.mozilla.cached-xul" mozilla will treat it just like any other "unknown" type. The only potential problem is a namespace conflict (if someone actually wanted to use that string for a real media type), which is not a problem now that it's in the vnd.mozilla subtree.
My concern only concern with the "cached" identifier is that, since it is entirely endemic to the application, it doesn't really fit what appears to be the application space of the MIME content type namespace. As I mentioned to Chris Waterson on the newsgroup, I am not sure how the registration authority will respond to this. If consistency with the MIME naming scheme must be maintained it *might* be preferable to use a type in the experimental subtree for this.
For the record, many of the things discussed here are relevant also to another quiet type, "text/xif" (for the XML Interchange Format that Mozilla uses internally).
What is the scope of that type, exactly? I mean, what would happen if a server sent content labeled as "text/xif" that was not the "xif" Mozilla is familiar with? If this is a non-issue because Mozilla would treat the content from the server as it would any other unknown type, it is probably better to rename the xif content type to something that won't impinge upon the MIME content type namespace. If, OTOH, this situation would pose a problem for Mozilla, then we need to see about registering a type for xif, again probably in the vendor subtree.
I have spoken with someone who has experience registering a type in the vendor subtree, and it appears that no formal documentation is required. Basically we just need to fill out this form: <URL:http://www.isi.edu/cgi-bin/iana/mediatypes.pl> I can do this, but let me run down my proposed entries to make sure there's consensus... 1. MIME Media Type Name: text 2. MIME subtype name: vnd.mozilla.xul-xml 3. Required parameters: n/a 4. Optional parameters: n/a 5. Encoding considerations: The encoding considerations for text/xml apply. 6. Security considerations: XUL shares security issues common to all XML content types. Furthermore, certain applications of XUL could potentially present the problem of "user interface spoofing", in which users might be presented with an interface requesting personal information and artificially designed to look like an application or system interface. The measures taken (if any) to minimize the possibility of such abuse are left to the implementation. 7. Interoperability considerations: n/a 8. Published specification: n/a 9. Applications which use this media type: the Mozilla internet application suite 10. Additional information * Magic numbers: n/a * File extensions: xul * Macintosh File Type Codes: [Can a Mac person comment on this?] * OIDs: n/a 11. Person to contact for further information: [I'll enter myself here if necessary, but it may be better to have someone who can speak as an authority on XUL. Dave?] * Intended usage: Common * Author/Change controller: mozilla.org
While I'm unsure as to whether or not XBL needs its own MIME type, it may be worthwhile to just register it anyway at the same time.
I don't see any harm in just crossing that bridge when we come to it; and proactively reserving MIME types doesn't strike me as the Right Thing. But if someone else feels otherwise, or when the need actually arises, a new bug can be opened to track that registration. Does the information I posted look accurate and acceptable? Should I list myself, Hyatt, or someone else as the contact person?
You might want to check out http://www.imc.org/draft-murata-xml. It looks like you've gotten the -xml suffix in place, if I'm reading Bugzilla correctly, but you probably want to use application/ rather than text/. The IETF has been unhappy lately with text/ types that aren't really meant for (ordinary) human consumption, technically 'casual readers'. -Simon St.Laurent (simonstl@simonstl.com)
Mass moving M18 bugs to M19
Target Milestone: M18 → M19
Ack, sorry for leaving this one hanging. Unless someone objects before Tuesday, 11 April, I will submit the MIME type form as specified previously, with only one (important) change: the type I intend to register is "application/vnd.mozilla.xul-xml", reflecting the suggestion of Simon St. Laurent.
Keywords: helpwanted
Sounds good to me. You probably want to mention this on the newsgroup as well so more people have a chance to object. Also, if no one objects, I will post a patch for that same content type -- "application/vnd.mozilla.xul-xml" -- on the same date. It will only address the former "text/xul" type; I'm not going to touch the "text/cached-xul" or other "internal" types (yet). That's a different issue.
Sounds good to me. You probably want to mention this on the newsgroup as well so more people have a chance to object. Also, if no one objects, I will post a patch for that same content type -- "application/vnd.mozilla.xul-xml" -- on the same date. It will only address the former "text/xul" type; I'm not going to touch the "text/cached-xul" or other "internal" types (yet). That's a different issue.
A little later than I'd said, but the application has been submitted. Here is a copy of it: Name : Braden McDaniel E-mail : braden@endoframe.com MIME media type name : Application MIME subtype name : Vendor Tree - vnd.mozilla.xul-xml Required parameters : Optional parameters : Encoding considerations : The encoding considerations for application/xml apply. Security considerations : XUL shares security issues common to all XML content types. Furthermore, certain applications of XUL could potentially present the problem of "user interface spoofing", in which users might be presented with an interface requesting personal information and artificially designed to look like an application or system interface. The measures taken (if any) to minimize the possibility of such abuse are left to the implementation. Interoperability considerations : Published specification : Applications which use this media : the Mozilla internet application suite Additional information : 1. Magic number(s) : 2. File extension(s) : xul 3. Macintosh file type code : ^@ Person to contact for further information : 1. Name : Dave Hyatt 2. E-mail : hyatt@netscape.com Intended usage : Common Author/Change controller : mozilla.org Dave, I went ahead and listed you are the person to contact for further info, as you seem to be the authority on XUL (as well as the owner of this bug). If that turns out to be inappropriate, it presumably can be changed. BTW, the message I got when submitting the application said "please allow at least two weeks for processing."
Okay, that's it. This patch plus the registration that Braden just submitted should do it.
Okay, registration has not succeeded; I got a note indicating that there is a problem with the application. Here is the pertinent fragment: > > Optional parameters : > > Since this is an XML-derived type, it needs to be clear whether or not > XML's charset parameter can be used, and if it can't, how charset > information is specified. Hyatt, can you or someone else who can speak authoritatively about XUL comment with regard to this?
mass-moving all bugs to m21 that are not dogfood+ or nsbeta2+ or nsbeta2-
Target Milestone: M19 → M21
Nominating nsbeta2+ 6/15. If we don't make that, should be nsbeta3 stopper. Here's why: defining the MIME type for XUL falls into the category of "things that must be gotten right by FCS, because it can't be changed after FCS and whatever we do for FCS we're stuck with for all eternity." It's not like a missing feature that can be pushed off to FUTURE, because system administrators all over the web will configure tens of thousands of servers to match whatever MIME type we use at FCS, and we'll be locked in to that MIME type forever. As documented in this bug report, the MIME type we currently use, text/xul, violates RFC2048. So this is a standards compliance bug, and I've already been getting questions on this (e.g. in my presentation at XTech 2000). Better to get it right the first time than to be hassled ever after. Good news is that there is already a patch submitted to implement the needed change. Resetting to M17 though it may slip to nsbeta3. cc:ing rickg as he mentioned that there would be a need to tweak the parser. Keith, does you patch take care of that issue as well? If not, would it be possible for you to take care off that as well per rick's instructions? We are stretched *very* thin right now and any further help you can offer would be greatly appreciated! Thanks to you and Braden for your help on this!
Keywords: nsbeta2, nsbeta3
Summary: MIME type for XUL invalid → MIME type for XUL invalid (not compliant with RFC2048)
Whiteboard: [HELP WANTED] → [HAVE FIX]
Target Milestone: M21 → M17
As mentioned, this bug is awaiting comment from someone who is an authority on XML. IMO, it doesn't make sense to apply any patches until we have a content type registered for XML. Based on previous experience, that will take 2-3 weeks from whenever the application is submitted--which I will gladly do again as soon as the aforementioned issue is clarified.
Braden, in the previous comment you say "XML" but I suppose you meant "XUL". I am not an authority on XUL, but if I may express my opinion, I think it would be best if XUL simply follows XML's lead in the character encoding area. Perhaps the XUL registration would then need to explicitly say that character encoding is indicated in the same way as XML itself?
Erik: Yes, I meant XUL. Sorry for the confusion. I agree that XUL ought to behave per the norm for XML documents--that it should use the character encoding parameter just as "raw XML" does. The aforementioned Internet draft ("XML Media Types") says (4.2): Optional parameters: charset Although listed as an optional parameter, the use of the charset parameter is STRONGLY RECOMMENDED, since this information can be used by XML processors to determine authoritatively the charset of the XML MIME entity. The charset parameter can also be used to provide protocol-specific operations, such as charset-based content negotiation in HTTP. "utf-8" (see [RFC2279]) and "utf-16" ([RFC2781]) are the recommended values, representing the UTF-8 and UTF-16 charsets, respectively. These charsets are preferred since they are supported by all conforming processors of [XML]. If an application/xml entity is received where the charset parameter is omitted, no information is being provided about the charset by the MIME Content-Type header. Conforming XML processors MUST follow the requirements in section 4.3.3 of [XML] that directly address this contingency. However, MIME processors that are not XML processors should not assume a default charset if the charset parameter is omitted from an application/xml entity. There are several reasons that the charset parameter is authoritative. First, recent WWW servers have been improved so that users can specify the charset parameter. Second, [RFC2130] specifies that the recommended specification scheme is the "charset" parameter. On the other hand, it has been argued that the charset parameter should be omitted and the mechanism described in appendix F of [XML] (which is non-normative) should be soley relied on. This approach would allow users to avoid configuration of the charset parameter; an XML document stored in a file is likely to contain a correct encoding declaration or BOM (if necessary), since the operating system does not typically provide charset information for files. If users would like to rely on the encoding declaration or BOM and to hide charset infromation from protocols, they may determine not to use the parameter parameter. Since the charset parameter is authoritative, the charset is not always declared within an XML encoding declaration. Thus, special care is needed when the recipient strips the MIME header and provides persistent storage of the received XML MIME entity (e.g., in a file system). Unless the charset is UTF-8 or UTF-16, the recipient SHOULD also persistently store information about the charset, perhaps by embedding a correct XML encoding declaration within the XML MIME entity. A pertinent question at this point: Does the current XUL implementation conform to this description?
I don't know. It would be good to get this tested. Any volunteers?
I'm not sure what sort of tweak would be necessary. My patch changes one line in htmlparser/src/nsIParser.h. It doesn't touch any of the actual logic. However, I have tested it and everything seems to work correctly. Perhaps the difference is subtle and I'm not noticing. Could someone please elaborate on how the parser is involved in this? If additional work is needed, I'm more than willing to help out as long as I have a clear understanding of what is required.
[nsbeta2-] won't hold NSBeta2 for this, but module owner could still apply this patch.
Whiteboard: [HAVE FIX] → [nsbeta2-] [HAVE FIX]
Status update, guys? Are we ready to land?
Braden, Keith: FYI, here's the situation: The Netscape PDT has marked this [nsbeta2-], which means that Netscape resources won't be expended to check this in. *However*, this doesn't prevent you as independent contributors from checking it in yourselves if you're willing to take responsibility for that (making sure you don't break the tree, dealing with any side effects, etc.). To do that, here is the procedure to follow IIRC: (1) get the permission of the relevant module owner for the checkin, and (2) get the permission of brendan@mozilla.org, who acts as the gatekeeper for independent contributor checkins during stabilization in the same way that the PDT acts as the gatekeeper for Netscape staff. I would strongly encourage you to follow this procedure and go through with the checkin. Thanks to your efforts, we're *this* close to fixing the bug and complying with the spec. If you can go just this bit of extra distance, we'll be all the way there. I hope you can drive this through and provide another great example of independent contributors improving Mozilla and increasing standards compliance! Thanks for all your help!
I have eliminated [HAVE FIX] from the status whiteboard, as it is misleading. None of the attached patches will fix this bug ON THEIR OWN. Until a MIME content type is actually registered, the patches will make us no better off than we are now. And until someone can answer my questions about XUL's handling of the charset parameter, the MIME type registration cannot proceed.
Whiteboard: [nsbeta2-] [HAVE FIX] → [nsbeta2-]
I'm adding jbetak@netscape.com to the Cc list. Juraj, I believe you own the so-called charset handling issues. Is that right? Please have a look at this bug report, which raises a number of questions regarding charsets in XUL and even in XML itself. Thanks.
nsbeta3-. not gonna do it.. wouldn't be prudent, not at this juncture...
Whiteboard: [nsbeta2-] → [nsbeta2-],[nsbeta3-]
Target Milestone: M17 → Future
Do what? Answer a simple question about the current functionality Netscape's provided Mozilla with? That's what's holding this up at the moment. Why is that so difficult? If there are bugs/missing functionality with regard to the charset stuff, those issues need to be filed. We just need some information here.
Braden: as far as we can tell, 'charset' is ignored for text/xul and not for text/xml. We probably intend to do it right for text/xul eventually, but won't for this release. For now, there is no way to change the charset, and it is always utf8. Does that answer your question or is there something else you need to know? Reassigning to Braden since he has voiced the intention of doing what remains to be done. ;-)
Assignee: hyatt → braden
Status: ASSIGNED → NEW
I've mailed Braden about this bug, but he hasn't replied. Is there any way this is going to get fixed before FCS, as it's one of those "if it's not, we're stuck with it for life" things, and it'll make us a laughing stock with the standards people... Gerv
like text/javascript? cc'ing the Dr.
Braden: Trudelle's "not gonna do it" comments in re: nsbeta3- was a mass bug change (you can find a bunch of others marked with the same comment on that date). I'm assuming his comments about "although it's minused, the module owner can still apply patches" still hold true. I think the short answer is that XUL currently only supports UTF-8, not due to any sort of design-based reasoning but because our XUL DOM and our XML DOMs are currently separate. With regard to the XUL /language/, it is XML and theoretically follows XML's rules about charsets. Of course, you have to be able to write the XUL tags in whatever charset you happen to pick, but XUL doesn't make any funny restrictions on what bytecodes it's looking for -- that's up to the XML parser, just like in any other XML application. In other words, XML's charset parameter /should/ be usable, and that's what you or we should tell the registrar. Please let me or David know what the status of registration is and we'll try to keep the ball rolling. I'll take this bug if you don't want to keep it. Re-setting component to "evangelism" since, if I'm not mistaken, that's reserved for this purpose.
Component: Parser → Evangelism
Claiming that using a MIME type of "text/xul" will make us a "laughing stock" is overreacting a bit I think. Let's not blow this out of proportion. :)
Sorry for the delay in my response; I've been very busy lately with work. Gervase: Mozilla already tacitly endorses the wholesale invasion of the IETF namespace for URI schemes. The commercial interests involved don't appear very interested in standards that aren't bolstered to sex-symbol status by the W3C, and mozilla.org management appears to be content with that. Thus Mozilla's commitment to standards support is already rather shakey, and I doubt that the outcome of this particular battle will stem the tide of abuse. Dan: Thanks for reminding me that there *are* folks at Netscape who give a damn about this. I'll resubmit the content type application. Note that the Internet draft "XML Media Types" has been updated, and the changes do impact the decision here. The current "correct" candidate would appear to be "application/vnd.mozilla.xul+xml". More importantly, the section I quoted before (4.2) nolonger exists. I suspect that the Right Thing is to act like application/xml in this regard; there are some examples in section 8 of the updated version that clarify correct behavior in this regard. I will follow up with a copy of the content type application.
Status: NEW → ASSIGNED
Here is the promised copy of the application. Scream if anything looks wrong. -- Name : Braden N. McDaniel E-mail : braden@endoframe.com MIME media type name : Application MIME subtype name : Vendor Tree - vnd.mozilla.xul+xml Required parameters : none Optional parameters : charset The qualifications regarding this parameter are the same as for application/xml. Encoding considerations : The encoding considerations for application/xml apply. Security considerations : XUL shares security issues common to all XML content types. Furthermore, certain applications of XUL could potentially present the problem of "user interface spoofing", in which users might be presented with an interface requesting personal information and artificially designed to look like an application or system interface. The measures taken (if any) to minimize the possibility of such abuse are left to the implementation. Interoperability considerations : XUL is not described by a DTD and applies only the well-formedness rules of XML. It should thus only be parsed by a non-validating parser. Published specification : n/a Applications which use this media : the Mozilla Internet application suite Additional information : 1. Magic number(s) : None. 2. File extension(s) : .xul 3. Macintosh file type code : "TEXT" Person to contact for further information : 1. Name : Dan Rosen 2. E-mail : dr@netscape.com Intended usage : Common Author/Change controller : mozilla.org
Dan: Do we need another bug filed to cover the handling of the charset parameter for XUL?
Sure. File that against me.
Braden: One of our wish-list items actually is to get a DTD for XUL written up. We're generally trying to do the Right Thing and encourage authors (through gentle force) to write valid XUL. Our roundabout way of doing this while developing the language was to "deprecate" XUL elements no longer intended for use by making them show up with a big offensive border around them using CSS. Anyway, If I were you, I would leave out the part about XUL only needing to be well-formed and not valid. That probably won't always be the case.
I doubt we can ever enforce any attribute restrictions on XUL (nor do I think we really should). People find the ability to attach arbitrary attributes onto an element for state-saving purposes to be very convenient, and this pattern has made it into widespread use in our XUL. I'd be reluctant to disable this capability, because it's so useful. e.g., mailnews does stuff like this... <treeitem mailtype="imapserver"> and you can then style the icon on the item from CSS [mailtype="imapserver"] { list-style-image: url(imapfolder.gif); } The RDF -> XUL transformation that is done through XUL templates also relies heavily on RDF properties being expressible on the XUL elements as attributes.
Also, using XBL you can make up your own new XUL tags. So in order to restrict the tag set, you would have to force any new tags to be placed into a namespace other than XUL. Right now people make up new tags using XBL and keep them in the XUL namespace, so even the tag set is variable.
David: Since XUL has HTML-style class selectors, I don't think there's anything you can do by adding attributes ad hoc that you couldn't do just by adding a class. So disabling Dan: David and I spoke about this on IRC. I don't think a DTD will ever be feasible for XUL. The main reason is that XML's evolution is slowly but steadily nudging DTDs into the realm of legacy practice. As it stands, namespace usage and DTDs are incompatible. The W3C's solution to the problem is schemas. I don't yet know much about schema-based validation, though. David mentioned the practice of adding attributes ad hoc to be used with attribute selectors in CSS. This is obviously incompatible with DTD-based validation, and I'd be surprised with schema-based validation accommodated it. The point is that the window to make XUL a more rigorous language is probably closing fast, if it hasn't already. Until there is an XUL schema and tools to apply it for validity checking, an interim solution might be a well-publicized "best practices" document, including a list of dos and don'ts. As far as the language in the content type application goes, we can submit an amendment any time. But I'm inclined to leave it as-is until we're clear that we actually want the additional rigor that validity will impose.
Braden - thanks for doing all this legwork. :-) Was the application successful? Gerv
I'm still waiting for a response.
The application has succeeded. application/vnd.mozilla.xul+xml is now an official MIME content type. Next step, then, is to make the appropriate changes to the browser source. Keith, are you still following this? And if so, are you interested in updating your patch?
Hey Braden: could you do a favor and attach the "success" email you received to this bug? It would be good to have that in a more permanent location for those interested than your inbox :) Also, I doubt this would be a candidate for something which could land in Netscape RTM or Mozilla 0.9 (not sure about that one, but blah blah blah risk and stuff :), but I have a strong feeling that it'll be very much desired for Mozilla 1.0. Thanks very much for your legwork on this, Keith and Braden!
Since Keith hasn't followed up, I've attached an appropriate patch. I think this one's ready to go now. Changing milestone to mozilla 0.9. Also, this is not an Evangelism issue; changing component back to Parser.
Component: Evangelism → Parser
Keywords: patch
Target Milestone: Future → mozilla0.9
At this point, I think it's too late to remove the old MIME type (since I know PDT wouldn't approve it) for NS6. Given that NS6 is going to ship with the old MIME type, we're going to need to continue support of it. So I'd prefer a patch that adds the new MIME type without removing the old one.
That would defeat the entire purpose of this bug report. Mozilla has no business arbitrarily invading the IETF namespace. Supporting "text/xul" *is part of the bug*. It is not a solution.
We could run it up the flagpole I suppose, but there's no way PDT is going to let this through.
Then if PDT doesn't accept this, they will be solely responsible for a future version of Mozilla being incompatible.
No they won't. If PDT doesn't accept it, Mozilla will support text/xul as well.
David: *Netscape* is responsible for a large portion of the delay in getting this bug fixed. *Netscape* decided this wasn't important enough to devote their own resources to. *Netscape* decided to base a product release on a codebase advertised as immature. And this bug is there in the first place because a Netscape engineer didn't bother to follow the proper protocol when introducing a MIME media type. For Netscape to attempt to shove this bug down mozilla.org's throat just because it's convenient for them is unreasonable and *extraordinarily* bad precedent.
If a distributor of Mozilla wishes to release software of pre-release quality then that is their decision. However, I don't think it is acceptable for the Mozilla design to be compromised by these releases. I believe it would be a bad precedent. If we don't even include an addition patch, is this going to jeopardise forward compatibility because people will want to target N6 and hence use the old type? Or can we expect XUL files will target a specific version of NGLayout? If the old type is not removed, I think it should at least be deprecated and release noted as going away in a future version, and to use the new type instead.
Pinging mitchell and brendan for input. This is an important policy issue, I believe.
The old MIME type can be deprecated but still supported. We can encourage people to use the new MIME type, but we can't simply drop support for it if NS6 ships supporting it.
Mozilla has been supporting text/xul for too long to drop that made-up type without a length deprecation/obsolescence cycle. I really have a hard time getting as worked up about this as others do, but I've been studying talkback data and trying to fix crashers lately. /be
s/length/lengthy/ /be
Oh, come on, Brendan. How many servers can you name that currently send the bogus type? At least be honest about this: mozilla.org is ultimately subject to the whim and requirements of Netscape even for what goes onto the trunk. And Netscape developers in positions of power are free to delay and ignore until time constraints "justify" Netscape getting its way by fiat. Well, I've had enough. You can find some other fool to jump through hoops for you. Reassigning to dr since he voiced some interest a little while back. I nolonger give a rat's ass.
Assignee: braden → dr
Status: ASSIGNED → NEW
Important policy issues don't just go away because there's a Netscape release happening. I guess it's OK for Mozilla to support it for a little while, but there has to a guarantee that it won't just have an "in future" status forever. If N6 doesn't support the new type, and the old type is not removed eventually, then people are just going to use the old type forever as far as I can see, because it works more often. Kind of like how people follow WWW un-standards because they work whereas up until Mozilla the W3C standards haven't. You do remember that situation don't you? The one you were trying to fix? And the one that has a lot of work ahead to solve? Why go backwards? Is this going to be nominated? It's essential we get the new type in at least. We'll see whether Netscape really gets it. You can cut features, you can leave in performance problems, but once you start selling out the future by leaving in bugs that prevent proper implementations in the future, you make enemies. The last thing that Netscape needs right now is enemies. They need evangelists, and they need web developers on their side. This sort of thing is exactly the sort of thing that has led us into the current mess, and I believe Netscape's future in part relies on it being a beacon of light. If you don't support a CSS1 feature, it can be added in a later version, even if it is a broken promise. This can't just be added later. You may think this is a trivial issue. Maybe it is. But it is indicative of an attitude that stinks. An attitude that has got us into this mess. An attitude I had hoped had gone. Has it PDT?
Well, let's find out. Nominating for RTM. Somebody needs to get review and super-review for the patch -- the fix is so straight-forward that the reviews should take all of five minutes each. PDT: This bug is more important than a crasher. Netscape 6 can be released with unfixed crasher bugs, without precluding the possibility of fixing them in the next point release. However, if Netscape 6.0 is released with this bug still open, that will effectively prevent the bug from being properly fixed (in Mozilla or Netscape) for *years*. The fix is nothing more than the replacement of one string by another, in the various places it occurs in the Mozilla source code -- 26 lines in total.
Keywords: review, rtm
Comments on the patch: I see a place where mailnews uses the MIME type as part of a contractID that looks like a GET request with name/value pairs. That makes me think they may be using it like that... I'm wondering if the + character in the new MIME type will cause problems there. Comments on your comments: Feel free to bring the bug before the PDT. If it can get into Netscape 6, then I'm happy to remove the old MIME type. The only reason I say we have to support it is that if we ship Netscape 6 with it, then - at the very least - Netscape 6.5 will have to support it. I suppose we could fork every file that contains the MIME type just to make sure Netscape continued to support it while Mozilla didn't, but that's hardly practical. As a basic rule, once you ship with something, you usually need to support it. You can deprecate the old MIME type, but you can't just throw it out. Sure, Mozilla could stand for principles and all and just throw out the MIME type, placing the burden on Netscape, but all that would mean is that people would write XUL pages for Netscape that Mozilla wouldn't be able to display. That said, if PDT is willing to take the patch for NS6, then it's ok IMO to kill the old MIME type. I'll give an a= once someone from mailnews (alecf or mscott) can verify that the + in the contractid won't hose anything they do with their converters.
>Oh, come on, Brendan. How many servers can you name that currently send the >bogus type? I honestly don't know, and that is one reason not to break those servers' clients (Zope comes to mind). Why is the burden on me rather than on you to find out who in the community might be affected for the worse by your patch? It's not my patch. >At least be honest about this: mozilla.org is ultimately subject to the whim >and requirements of Netscape even for what goes onto the trunk. Be honest yourself. If you want me personally to go land your patch in the trunk, breaking who-knows-what part of the community that uses remote XUL, you are looking for a free lunch. "Netscape" is a big organization, and most of the people in it don't know or care about this bug. Stick to individuals if you want accountability. I just re-read this bug, and there is no "whim" here except for hyatt's original choice of text/xul based on ignorance of the rules. After that, I see nothing but slow movement toward the current patch. >And Netscape developers in positions of power are free to delay and ignore >until time constraints "justify" Netscape getting its way by fiat. Are you claiming that someone (erik? hyatt? dr? who?) delayed and ignored your question about charset (which you could have answered for yourself with a little code inspection)? That those evil people did so intentionally so they could get away with inflicting their whim on the world in Netscape 6? That's rubbish! I don't care about what the PDT does in Netscape 6, because mozilla.org is not responsible for that product or its branch, but before anyone lands this patch on the trunk, they had better ascertain how many people in the Mozilla community depend on text/xul, and deprecate before obsoleting the old type. At least ask in all the relevant newsgroups, contact the Zope people, do some leg work! I've been consistent in saying this, and it has nothing to do with Netscape or their whim-conspiracies, so please give that a rest. /be
cc'ing alecf and mscott per hyatt's last comment. hyatt: it's not ok to dump the old type on the trunk without a well-publicized warning followed by a decent interval before doing the deed, if not by a full Mozilla milestone deprecation period. /be
Brendan, understood.
Based on brendan's comments, I need a new patch that adds the new MIME type without removing the old MIME type. I'll review it when it gets attached.
> I honestly don't know, and that is one reason not to break those servers' > clients (Zope comes to mind). Why is the burden on me rather than on you to > find out who in the community might be affected for the worse by your patch? > It's not my patch. I had a longer, more thorough response prepared; but you know, it all comes down to this, so I just snipped it: Which Mozilla APIs have been announced as frozen? Because I wasn't aware that this patch tread on any that had. I thought I was still operating in the realm of unreleased, in-flux software.
>Which Mozilla APIs have been announced as frozen? Because I wasn't aware that >this patch tread on any that had. I thought I was still operating in the realm >of unreleased, in-flux software. Nothing is frozen, despite attempts to freeze a tiny set of APIs. What does that have to do with anything? If you break Zope, mozilla.org will hear about it. I'm trying to head off that avoidable bit of trouble by holding you to the same standards you demand of netscape.com people: announce first, discuss and get consensus, then make an incompatible change ("then" means soon, or after a major deprecation release). Believe it or not, there are other companies out there than netscape.com that are shipping Mozilla code. We were even compelled last year, by complaints from sun.com and by the size of their user base, to "freeze" parts of the plugin API because JRE1.3 went out with binary dependencies on that API. (Lest someone change the subject to deflect attention from this bug, yes -- it sucks that mozilla.org doesn't know exactly who is shipping product based on versions from our respository. We've reached out several ways to enquire about who might be doing that, but if you have better ideas for how to find out, mail me. We've already tried the "Mozilla 1.0 is next year, please wait" line, but it's a free world and people will use this year's Mozilla at their discretion.) /be
de-nominating for rtm, and cleaning up other spurious keywords... this won't be "fixed" to everybody's satisfaction until the deprecation period for text/xul passes, at which time we can claim well-behavedness when we remove it for good. the patch needs to be worked on to add support for both mimetypes, which i suppose i'll do (since braden seems to have nominated me). i'll even try to get my patch into rtm, although i don't believe any of us have a convincing argument to that end (i'll do this in a different bug, and keep this one open until, as braden asserted, the made-up mime type is no longer supported). a couple comments, though... braden: for someone who "doesn't give a rat's ass," you're still fighting pretty hard. and don't get me wrong -- i have plenty of respect for that -- but understand that there is precedent even within the standards bodies of abusing other standards. the w3c themselves managed to pull the "text/javascript" pseudo-mime-type out of their ass in the html4 spec, which isn't even an appropriate type (it ought to be application/javascript). consider also the example of xml namespace uris: they're not even regulated, and although people are bouncing ideas around for how to do so, anybody at any point in time could trounce anybody else's namespace. i'm sure you can find more examples of this sort of thing if you look around... regardless, i'd prefer not to get in on this brawl. i'll try to fix things to the best of my ability -- that's the best i can promise.
Status: NEW → ASSIGNED
Keywords: helpwanted, patch, review, rtm
Target Milestone: mozilla0.9 → mozilla1.0
> Nothing is frozen, despite attempts to freeze a tiny set of APIs. What does > that have to do with anything? Well, gee, Brendan, what on earth *could* that have to do with *anything*? I wonder why that might lead a contributor to believe that the product is in flux, and persons using the code are using pre-release code *at their own risk*? Maybe the fact that this isn't true should be more well-publicized by mozilla.org. And please keep my comments in context when responding to them. The response I got to my patch was not, "No, we have too many potential *existing* dependencies," it was, "No, because there's a major Netscape release coming up and PDT won't approve this." For you to pretend that I'm just miffed because I'm being asked to do what you perceive as due diligence is disingenuous. > If you break Zope, mozilla.org will hear about it. I'm trying to head off > that avoidable bit of trouble by holding you to the same standards you demand > of netscape.com people: announce first, discuss and get consensus, then make > an incompatible change ("then" means soon, or after a major deprecation > release). Excuse me? I don't have any problem with netscape.com persons making incompatible changes to unreleased code as long as it's clear that the changes are improvements. Objections I might have raised in the past have been to things which weren't necessarily clearly improvements, and that covers most any significant architectural change. This is not an architectural change. And unless I've been totally misled about mozilla.org's stance on standard conformance, this is a clear improvement. I know how convenient it might be to portray me as a hypocrite here, but it's not gonna wash. > (Lest someone change the subject to deflect attention from this bug, yes -- it > sucks that mozilla.org doesn't know exactly who is shipping product based on > versions from our respository. We've reached out several ways to enquire > about who might be doing that, but if you have better ideas for how to find > out, mail me. We've already tried the "Mozilla 1.0 is next year, please wait" > line, but it's a free world and people will use this year's Mozilla at their > discretion.) Indeed. "At *their* discretion". That phrase doesn't really convey to me that *Mozilla contributors* are obligated to go on wild goose chases to hunt down anyone who might be affected by a patch. I'm not fundamentally against doing more legwork here. But, Brendan, the rationale you've presented for requiring that legwork is woefully inadequate. I don't see how you can acknowledge that clients are currently using unreleased, unfrozen Mozilla code "at their discretion" and simultaneously claim that Mozilla contributors are responsible for making sure those clients don't take collateral damage from patches. It doesn't make sense.
Keywords: helpwanted, patch, review, rtm
Target Milestone: mozilla1.0 → mozilla0.9
Sorry about overwriting your keyword changes, dr. The alternative was losing a long comment, and I didn't want to do that. I think I've put things back as you had them.
Keywords: helpwanted, patch, review, rtm
Target Milestone: mozilla0.9 → mozilla1.0
dr: I guess I care in spite of myself. I don't care much about the content type anymore, but this is a helluva lot bigger than that now.
braden: regarding your comments to brendan -- nobody has a hard and fast obligation to make the mozilla platform users' lives easier. but you can't go tearing the carpet out from under their feet all in the name of standards. the moral obligation we have, regardless of opinion, is to strike a fair balance for everybody's benefit. i am a firm supporter of open standards, but even the standards bodies realize that they have to be pragmatic as a part of process. that's where learning, refinement, and improvement happen. anyway, that was a vaguely long-winded way of saying that we're doing the only reasonable thing to do here. please take further discussion to email.
>Well, gee, Brendan, what on earth *could* that have to do with *anything*? I >wonder why that might lead a contributor to believe that the product is in >flux, and persons using the code are using pre-release code *at their own >risk*? You don't get it still? I'll try again: module owners and patch owners should cooperate to make a change that breaks compatibility in this way public, gather feedback, and work out an agreeable way to land the change *before* checking in the change. Clear? >Maybe the fact that this isn't true should be more well-publicized by >mozilla.org. No, I think roadmap.html publicizes things well enough. I'm not going to revise it to say "look out, someone may rip out text/xul or a similar protocol-born type without any notice". >And please keep my comments in context when responding to them. I wasn't responding to your comments to hyatt -- you and hyatt work out your differences without me, please. I'm reminding you how to make incompatible changes in the Mozilla trunk, for the third time. Why am I here anyway? Do you really need me to help everyone cooperate? >The response I >got to my patch was not, "No, we have too many potential *existing* >dependencies," it was, "No, because there's a major Netscape release coming up >and PDT won't approve this." Hyatt seems to have changed his mind, and was willing to "run this up the flag pole". dr seems less optimistic. But as I wrote, that's all netscape.com's business. And (again, so you don't miss it) I'm not saying "No, never" -- I am not objecting to your patch going into the trunk simply because we have unknown (if not too many) dependencies on it. If you and the owners of the dependencies can come to agreement fast, great. To do that, you have to publicize the change in better ways that commenting in this bug. >For you to pretend that I'm just miffed because I'm >being asked to do what you perceive as due diligence is disingenuous. I didn't pretend any such thing. You attacked my comments asking for due diligence. I did take you to task for your ludicrous complaint that "Netscape" was delaying on purpose to impose the "whim" of text/xul on the world, but maybe you got over that silliness. I also didn't say you were a hypocrite. And I never said one word against the patch; in fact I am in favor of it going into the trunk as soon as is agreeable to affected parties. I'm sorry you think that it doesn't make sense to expect patch purveyors, owners, and users to satisfy or break _de facto_ standards and commitments in the Mozilla code by proposing an incompatible change outside of the confines of a bug like this *before* breaking the standards. I don't think that requirement is at odds with Mozilla being in an unfrozen state. You don't freeze a project this big all at once, and you don't break parts of it that have been checked in for a year or more, and on which others to whom you haven't spoken depend. Another parochial _de facto_ standard example: Mozilla has a bunch of crucial code implementing the not-yet-standardized XBL, used by many in the community, and it is getting hard for hyatt to make incompatible changes at all, never mind make them without advice and consent. I'm done here. /be
This may sound silly, but can't we recognize both content types as xul: text/xul and "appliation/vnd.mozilla.xul+xml" At least in the interim. That way we would still properly recognize servers that issue text/xul content as xul while also adding support for the more standards friendly content type as well.
That's what I suggested, but braden and matty are objecting to this idea, asserting that any support for the text/xul MIME type constitutes a violation of the standard.
triaging for post-rtm
Target Milestone: mozilla1.0 → mozilla0.9
A better solution all round would be to use text/xml as the MIME type for XUL, and fix our content sink / parser / network library to instantiate the document object only after finding out what the namespace of the root element is or what the DOCTYPE of the document is.
Dave, it would be in violation of the standard, but I said it would be OK IMO to merely deprecate it for now if it's going to be removed eventually. But this won't be possible ever unless N6 supports the proper type. Brendan, maybe you should think seriously about breaking released products if you need to. It might make others think twice about not being in contact with mozilla.org. Matthew, I agree, this is more serious than a crasher bug, at least the addition part.
Braden mailed me about my being less than clear in commenting about how long a deprecation period was required, before obsoleting text/xul. My comments here were my opinion, vaguely expressed at first using "lengthy" to mean no longer than one Mozilla major milestone (three months, mozilla0.9). But I went on in subsequent comments to write that with discussion, it might turn out that no one was much put out by removing text/xul; or that people could cope more quickly than my unspecified-till-now "lengthy" implied. I was not laying down mozilla.org policy on deprecation and obsolescence of MIME types or other protocol-born data on which distributed consumers of Mozilla may have come to depend. I don't believe mozilla.org can devise a general policy at this time. I was only giving my vaguely-worded opinion in my first comment, when I should have been clearer and gotten on to the "discuss among yourselves in a public newsgroup, I'm verklempt" state. Sorry about that (and thanks for pointing my moment of unclarity out to me, Braden). /be
Okay... I admit it... I care. I *really* wish I didn't, but I do. Per Brendan's request, I've started a thread on the .xpfe newsgroup. I'll e-mail Martijn Pieters about this thread. If any of you know of other folks who ought to be alerted about this, please do so.
upon refactoring / rewriting of xml content sink, this bug will be fixed per ianh's recommendation: text/xul will be deprecated in favor of text/xml, and the application type will be determined by namespace rather than mime type. after a release cycle or other suitable period of "warning" time, support for text/xul will be removed. targetting for moz1.0 (contingent on xml content sink work).
Summary: MIME type for XUL invalid (not compliant with RFC2048) → deprecate and remove non-compliant text/xul mimetype.
Target Milestone: mozilla0.9 → mozilla1.0
Sounds reasonable. But you want "application/xml", not "text/xml".
The sooner we sort this, the less painful it will be... Gerv
Keywords: mozilla1.0, nsbeta1
Totally. Once the nsXMLContentSink is reworked, we can use application/xml, and use the XUL namespace on the root <window> element to detect XUL documents. There may be an interim solution (read: ugly hack to nsXMLContentSink) but I'm not sure... Cc'ing jst -- Johnny, can you tell me what's going on with the content sink? Thanks.
Looking at the length of the discussion in this bug I won't go into any level of detail about the XML content sink here. As far as I know there won't be any dramatic changes to the content sink any time soon, there will probably be a major rewrite of the XML/XUL/XSL sinks but I don't see that happening in time for mozilla 1.0. Currently the document is created before the sink and that makes it really complicated to look at the namespace on the root node to determine of what type the currently loading document should be, and I don't see that changing before mozilla1.0 either.
->future, unfortunately. jst: if you happen to have a bug on you for the content sink work, do me a favor and make that block this. thanks.
Target Milestone: mozilla1.0 → Future
Dan, there's no bug on file for the content sink work yet, we're still just talking about it...
So, in summary: We are currently using text/xul. We eventually want to replace that with application/xml and work out that it's XUL another way. However this will take a long time, because it depends on content sink refactoring. So, in the mean time, do we want to do a search-and-replace to use the (legal) MIME type appliation/vnd.mozilla.xul+xml ? Or does that just make more trouble? I think we should avoid using text/xul a second longer than we have to... Gerv
please leave this alone. Afaik we can not do the replace because no one checked in support for that type.
Keywords: nsbeta2, nsbeta3
Whiteboard: [nsbeta2-],[nsbeta3-]
Can we not "check in support for that type" then? Surely the resolution is to replace all instances of text/xul in the entire Mozilla source with vnd/mozillablahblah ? How could it be more complicated than that? Gerv
Timeless, I thought you'd be the last person I'd have to tell that mozilla.org has made precisely zero major releases of its software, it's pre-release, and interfaces, UI, APIs and everything else are subject to arbitrary change without warning. I see no reason why potential future work, with a possible "better fix" for this problem should prevent us doing the right thing now - especially as the potential future work will not happen before our first major release. As a side note, how does changing our internal MIME type for XUL affect anyone apart from us? No-one's serving XUL up from a webserver to be interpreted in the browser, are they? I would have thought Security would have something to say about that... Gerv [ Despite what the Bug Activity might say, I have been following this bug for the past six months or so. ]
> mozilla.org has made precisely zero major releases of its software, Bugzilla is nearly at 2.12, and is planning a 3.0 for before i my sister graduates from college [4 years, at 16 web years should be plenty, please] > it's pre-release, and interfaces, UI, APIs and everything else are subject to arbitrary change without warning. This is wrong. Roadmap and brendan will tell you that we do freeze [or try to freeze] portions of apis at specific milestones (I can tell you that UI _did_ freeze on me at specific milestones), and we do provide warning on www.mozilla.org and in the newsgroups. > No-one's serving XUL up from a webserver to be interpreted in the browser, > are they? they are. people.netscape.com does, bugzilla.mozilla.org does, and at least a few other real groups, some day i might try to find a useful list of such sites, but i haven't figured out what to tell google when i ask. I see no reason not to accept the patch that supports both, except i'm not sure it exists. Offering a migration path is better than bickering and stikcing w/ a path that people don't like [although i don't mind the current path]
> Bugzilla is nearly at 2.12, and is planning a 3.0 for before i my sister OK, so I've been out-pedanted :-) Currently, the number of people who are serving as text/xul who are not under our direct or indirect control is very small. If we release Mozilla 1.0 with text/xul, not only do we look fools in front of anyone who cares about standards, but then we have to support the wretched thing for another major revision, or however long it is. If we don't change now, we have to live with this. Anyway, you and I arguing about it probably won't change much. I know Brendan is CCed on this bug, and he's the man to make the final decision. Brendan? Gerv
I don't think we should release mozilla1.0 with text/xul polluting the text/ MIME type space. We've been over this at some length, here and in mail/news messages. I'm expecting dr (or someone else motivated who'll take this bug) to do the .xml with a namespace thing in time for 1.0. Hence the milestone keyword nomination. Now, dr: why is this bug Futured? Are you using that milestone to mean "not by mozilla0.9"? How about finding someone else to do it in the current milestone (0.9), or sooner than at the last minute (during 1.0)? /be
This is excessively irritating. To detect that an XML file (application/xml) is XUL, the file needs to be at least partially parsed, if not by an XML parser then by some other parser which can "take a good guess" based on the presence of the XUL namespace in the document. To do this correctly, the XML content sink needs to be refactored to handle multiple strategies per-namespace. I am not willing to lean on jst to do this right now, because the XML/DOM team has bigger fish to fry at the moment. The other solution, simply replacing the incorrect uses of text/xul with the obtuse "application/vnd.mozilla.xul-xml" would be technically correct. It is also true that only a small number of authors or servers are serving content with the type text/xul currently. But I think we agreed long ago that switching content types for XUL without a "grace period" is inappropriate for many reasons. Some of these reasons involve mozilla.org vs. Netscape politics, which I think are well-understood by everybody here, and some of these reasons are purely technical. One major reason I believe everybody here should agree on is that the small core of people and organizations serving up XUL currently represents the segment of mozilla's user base which is reasonably well-versed in our technologies, and we don't want to piss them off. Regardless, the ground has been well-covered. I personally believe the right solution to be to improve or refactor the XML content sink to deal with pluggable strategies per-namespace. I would be willing to do that work, but it would both trample on the feet of that code's owners, be irrelevant to XPToolkit's short-term goals, and take a long time since I'm not at all familiar with the relevant code. I should also say that I personally believe the *wrong* solution to be a lengthy session of bickering. I'm reassigning this bug to nobody@mozilla.org at this point for the reasons I described, in the hope that somebody will take this bug and do the right thing. Perhaps this bug could be reassigned to the default owner of XML, or some such... I'd be more than happy to accept this bug again if it is agreed that the content sink work or some reasonable equivalent should be done. Otherwise I want no part in it.
Assignee: dr → nobody
Status: ASSIGNED → NEW
Is there a reason that a patch to support both "text/xul" and "application/vnd.mozilla.xul+xml" could not be checked in ASAP, and then another patch to remove support for "text/xul" checked in at some point prior to 1.0? That should be pretty trivial to implement, and I think I could provide such patches.
Braden: go for it. It would be better, I think, to work on the namespace-based strategy flexibility that dr describes, but we don't want text/xul in 1.0, so if we can't have no mime type for XUL, we should have a proper one. /be
Okay. I've filed bug 69749 on the dual support issue. I'll add that I agree that namespace-oriented pluggability is the Right direction to go for a number of reasons. Unfortunately, it involves enough work that it appears unlikely to provide a timely solution to this problem. And as has already been pointed out, timing is important here.
Assignee: nobody → braden
Depends on: 69749
Summary: deprecate and remove non-compliant text/xul mimetype. → remove non-compliant text/xul mimetype.
Target Milestone: Future → mozilla1.0
Status: NEW → ASSIGNED
QA Contact: janc → bsharma
updated qa contact.
Hyatt and Ben Goodger have decided to make backwardly-incompatible changes to the XUL syntax before 1.0. Would this not be an ideal opportunity to change the MIME type while we were at it? After all, two incompatible formats with the same MIME type is bad, and we've been trying to ditch text/xul for ages now... Two birds with one stone? Braden? Brendan? Gerv
Gerv: Are you suggesting skipping a deprecation period where both types would be supported? Or are you just advising some coordination in the chronology of the changes? The latter seems obviously Good; I've no opinion on the former. FWIW, I'd planned on posting a patch to bug 69749 either tonight or tomorrow. It touches a lot of files and should probably be reviewed by a fair number of people.
Braden: I'm suggesting skipping the deprecation period of text/xul because we don't want post-change Mozilla trying to read pre-change XUL anyway (because it'll barf.) This is a good (the ideal) way of making sure that doesn't happen. Hyatt, Ben G: do you have a view on this? Because that would be reasonably conclusive ;-) Gerv
Okay; I'll wait for a conclusion on this before taking any action on bug 69749. (Presumably, the namespace identifier for XUL will be changing as well?)
Braden: Hyatt said "OK" to me on IRC, for changing the MIME types (not the namespace) as detailed in this bug. Wahey! But because the XUL changes have just landed, we should do this ASAP. Are you going to prepare the patches? <looks hopeful> If not, let me know, and I'll try (though I've never done anything this complex before.) Gerv
OK - patch being attached. This changes text/xul to application/vnd.mozilla.xul+xml. It also changes text/cached-xul to application/vnd.mozilla.cached-xul . The reasoning behind this is that, even if we haven't registered it as an official type yet, we are being less horrifically polluting of the MIME namespace this way, and if we don't do it now, it'll never get done. Looking for review and super-review. Gerv
Gerv: Do not add "application/vnd.mozilla.cached-xul" (or any other MIME type outside of the experimental namespace) until you have registered it. Period. It creates bigger problems than it solves. (Bugs like this one, for instance.) I thought the conclusion was that "text/cached-xul" does not need to be changed since it is completely internal to the application. If we do want to change it, it should be either to something in the experimental namespace ("application/x.vnd.mozila.cached-xul") or, better, something in an application-specific format that would not be mistaken for a MIME type.
Yeah, cached XUL is not a real exposed MIME type. It's just a hacky thing we use internally. I don't really mind if it is changed though. sr=hyatt
Braden: If it's for internal use only, then it doesn't matter. If it's external, the new one is better than the old - because we don't have to change it _again_ if we later decide it needs registering. Do you think we _couldn't_ get it registered if we bothered? Whereas we wouldn't have a prayer of getting the old one. Gerv
Gerv: You *might* be able to get a "cached-xul" registered, but you shouldn't. It is not part of public interchange, therefore there is no reason to clutter the media type namespace with it. For that reason, "application/vnd.mozilla.cached-xul" is not significantly better than "text/cached-xul". Whatever string is chosen for this, *it does not correspond to a MIME type". If people get warm fuzzies from making it look like a MIME type (though personally I think this just incites confusion), then either use a name in the experimental namespace or use something that is more obviously application-specific like "mozilla.application/cached-xul". In any event, there is no good reason for "repairing" "text/cached-xul" to be part of this bug. It does not depend on the string used for the public XUL MIME type; and as an internal type it does not have the problem with external dependencies (which has been the principal factor in delaying resolution to this bug).
OK, you win. mozilla.application/cached-xul it is. Blake or Ben - still looking for an r=, and a swiftly-following check-in :-) Gerv
Keywords: mozilla0.9, patch, review
Done and done.
Status: ASSIGNED → RESOLVED
Closed: 26 years ago → 25 years ago
Resolution: --- → FIXED
I am adding support for text/xul back, so as not to break all those evil bad naughty sites that actually serve it up. Like mine.
This caused bug 74502. Apparently, somebody in the ``real world'' is serving up text/xul.
waterson: anyone who is serving text/xul will be serving broken XUL anyway, surely? What has happened here is that the change has uncovered bugs in Mozilla. Why is this different to any other time where this happens? We should fix the bugs, not back out the change. We fully expected there to be a few sites serving up XUL. These people would have had to have known about this change because they would have had to know about Blake's other changes. The only reason I haven't made an announcement yet is that I was waiting for bug 74146, which people seem to be ignoring, to get fixed. Gerv [Putting this comment in two places]
QA Contact: bsharma → moied
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: