1120171 - Remove support for moz-prefixed XHR responseTypes (moz-blob, moz-chunked-text, and moz-chunked-arraybuffer)

Status: RESOLVED FIXED

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

Description

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Should probably warn first.

Comment 1

[Kohei Yoshino]

Posted the site compatibility doc: https://www.fxsitecompat.com/en-US/docs/2015/prefixed-xmlhttprequest-response-types-will-be-removed/

Comment 2

[Yury Delendik (:yury)]

The PDF.js project is using moz-chunked-arraybuffer a lot. There is no Fetch and Stream API alternative API yet AFAIK. Shall we block this on bug 1147061 or bug (if exists) that will implement Streams API support for fetch()?

Comment 4

[Jaap Stolk]

https://developer.mozilla.org/nl/docs/Web/API/XMLHttpRequest does not (yet) indicate these are considered to be removed.

I'm doing server-push (long polling), and sending multiple parts of binary data by doing FCGX_FFlush() on the server and then handling the data on the XMLHttpRequest onprogress event in the client.
The remaining response types do not allow access to the received data, only indicate the number of bytes received. Is there a working alternative once the moz-prefixed types are removed?

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

MDN doesn't indicate that yet since we need first some telemetry data whether they can be removed
(bug 1335365).

Could you use WebSocket or EventSource?

(last time I looked at, fetch() wasn't too good for this)

Comment 6

[Jaap Stolk]

I would love a more portable solution, but have not found one yet.
I looked into WebSockets, but then need to compile (and maintain) a patched version of lighttpd. EventSource is included in lighttpd but as far as I can see is not suitable for binary data.
(For completeness, I'm using the moz-chunked-arraybuffer type, and an FCGI C-program on the server.)

Comment 7

[Anne (:annevk)]

I think for moz-chunked-arraybuffer usage fetch() is actually what you want, but we'd need to implement and ship bug 1128959 first. Doing that before removal seems reasonable to me.

Comment 8

[Jaap Stolk]

Will fetch() work when my C back-end is connected to the lighttpd webserver using FCGI on Linux? If these things are not clear yet, I may as well wait for support for webSockets in lighttpd.

Comment 9

[Anne (:annevk)]

fetch() is basically the new XMLHttpRequest, an API for doing HTTP requests and responses with some extras. And since your setup works with the latter, it'll work with fetch().

Comment 10

[Jaap Stolk]

Thank you, that is very interesting and also looks more portable. I'll see if I can use fetch() instead.

Comment 11

[Kohei Yoshino]

Updated the site compatibility note: https://www.fxsitecompat.com/en-CA/docs/2017/prefixed-xmlhttprequest-response-types-including-moz-blob-are-no-longer-supported/

Comment 12

[Andrew McCreight [:mccr8]]

A subtest of wpt/xhr/historical.html checks that moz-chunked-arraybuffer does not exist, and this fails, of course, only in Firefox.

Comment 13

[Thomas Wisniewski [:twisniewski]]

Yury, does PDF.js still need -moz-chunked-arraybuffer, now that ReadableStreams should be working? Should we try removing it now?

Comment 14

[Thomas Wisniewski [:twisniewski]]

baku, it seems that pdf.js is no longer a concern as far as removing platform support for -moz-chunked-arraybuffer (as per bug 1411865 comment 5). Did we know of any other reason to not try removing it, now that we support ReadableStreams?

Comment 15

[Andrea Marchesini [:baku]]

(In reply to Thomas Wisniewski [:twisniewski] from comment #14)

baku, it seems that pdf.js is no longer a concern as far as removing platform support for -moz-chunked-arraybuffer (as per bug 1411865 comment 5). Did we know of any other reason to not try removing it, now that we support ReadableStreams?

No other reasons. I still see browser/extensions/pdfjs/content/PdfJsNetwork.jsm using moz-chunk-arraybufer here:
https://searchfox.org/mozilla-central/rev/56705678f5fc363be5e0237e1686f619b0d23009/browser/extensions/pdfjs/content/PdfJsNetwork.jsm#102

If this can be removed, we can remove the support for moz-chunk-arraybuffer.

Comment 16

[Thomas Wisniewski [:twisniewski]]

In that case I suspect that we just need someone to review the patch on bug 1411865 so it can land.

Comment 17

[Rob Wu [:robwu]]

"moz-chunked-arraybuffer" usage is conditional on the onProgressiveData callback being set.
In mozilla-central this callback is never set: https://searchfox.org/mozilla-central/search?q=onProgressiveData&redirect=false

The callback is only set when used in the generic web version of PDF.js: https://github.com/mozilla/pdf.js/blob/9b5a937f78418ad6b114ca66d062cf2ae416d05f/src/display/network.js#L321-L322

So removing "moz-chunked-arraybuffer" will not affect the built-in PDF.js. At most it affects old PDF.js versions on the web that haven't been updated to use the streams API. In this case, the library will fall back to Range requests. Although not optimal, the basic functionality still works, so that shouldn't block the removal of this.

Comment 18

[Thomas Wisniewski [:twisniewski]]

Attachment: Bug 1120171 - remove support for XMLHttpRequest moz-chunked-arraybuffer response type; r?baku

remove support for XMLHttpRequest moz-chunked-arraybuffer response type

Comment 19

[Thomas Wisniewski [:twisniewski]]

A try run seems fine: https://treeherder.mozilla.org/#/jobs?repo=try&revision=0391a9b1bf367f6f78e8684438583e83cf5780f4

Comment 21

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/3ff9e491c530
remove support for XMLHttpRequest moz-chunked-arraybuffer response type; r=baku

Comment 22

[Raul Gurzau (:RaulG)]

https://hg.mozilla.org/mozilla-central/rev/3ff9e491c530

Comment 23

[Kohei Yoshino]

Posted site compatibility note: https://www.fxsitecompat.com/en-CA/docs/2019/xhr-moz-chunked-arraybuffer-response-type-is-no-longer-supported/

Comment 24

[Rob Wu [:robwu]]

The alternative to the functionality moz-chunked-arraybuffer is not arraybuffer, but using fetch and readable streams.

• XMLHttpRequest + responseType = "moz-chunked-arraybuffer" offered the ability to receive the response in chunks (as an ArrayBuffer; other types are available as moz-chunked-text and moz-chunked-blob, but the latter are not documented on MDN).
• XMLHttpRequest + responseType = "arraybuffer" offers the response as an ArrayBuffer, but only once the full request has been received.
• fetch + readable streams offers the ability to read the response in chunks - see https://developer.mozilla.org/en-US/docs/Web/API/Streams_API/Using_readable_streams#Consuming_a_fetch_as_a_stream

Could you update the webcompat article?

We should also update the "moz-chunked-arraybuffer" section of https://developer.mozilla.org/en-US/docs/Web/API/XMLHttpRequest/responseType , to mention the deprecation and link to the alternative.

Comment 25

[Kohei Yoshino]

Hi Rob, thanks for your feedback! I’ve corrected the site compatibility note accordingly.

Comment 26

[Rob Wu [:robwu]]

Looks good. I've also updated the MDN article above to note the removal and the alternative.

Comment 27

[Eric Shepherd [:sheppy]]

Rob, thanks for that update. I've made a few adjustments and have submitted BCD PR 4349 to update the compatibility data for this.

Also added to Firefox 68 for developers.