441496 - Use server push (multipart/x-mixed-replace) with Safari

Status: VERIFIED WONTFIX

(Bugzilla :: Query/Bug List, defect)

Description

[Gervase Markham [:gerv]]

As noted in bug 188712, It seems Safari has supported multipart/x-mixed-replace since 2.0.2 in 2005:
http://webkit.org/blog/32/webkit-fixes-in-safari-202-mac-os-x-1043/

We should remove the check which prevents Safari getting the intermediate page.

Gerv

Comment 1

[Gervase Markham [:gerv]]

Attachment: Patch v.1

Comment 2

[timeless]

Comment on attachment 326442 [details] [diff] [review]
Patch v.1

sorry. sniff for webkit 2+. i use 10.3.9 which means i can't run safari 2, which means you're going to break me.

but i do appreciate the review request and will grant a positive review for one which properly handles the distinction :).

Comment 3

[Gervase Markham [:gerv]]

Attachment: Patch v.2

OK, for those still using Safari 1...

Gerv

Comment 4

[Frédéric Buclin]

Comment on attachment 326443 [details] [diff] [review]
Patch v.2

>+$ENV{'HTTP_USER_AGENT'} =~ /WebKit\/(\d+)/;
>+my $webkitversion = $1;

If the UA doesn't match this regexp, I suspect $1 would still contain any previous result from an older regexp (I cannot test right now because I'm not on my PC). You should rather write

my $foo;
if ($ENV{...} =~ /.../) {
    $foo = $1;
}

to be safe.

Comment 5

[Gervase Markham [:gerv]]

Attachment: Patch v.3

Comment 6

[Gervase Markham [:gerv]]

Attachment: Patch v.4

Slightly embarrassing that this has taken four goes...

Gerv

Comment 7

[Gervase Markham [:gerv]]

Attachment: Patch v.5

Words fail me.

Gerv

Comment 8

[Frédéric Buclin]

Comment on attachment 326702 [details] [diff] [review]
Patch v.5

Looks good. r=LpSolit assuming you tested it with Safari.

Comment 9

[Gervase Markham [:gerv]]

I can't test it with Safari; I don't have a Mac. Do you?

Gerv

Comment 10

[Max Kanat-Alexander]

Pyrzak has a Mac! :-)

Comment 11

[Bob Williams]

I can confirm that it works with Safari 3.1.1 (5525.20) on my own Mac. I patched buglist.cgi, and search results still came up just fine. Note, however, that I have few enough bugs that I don't actually see the intermediate page, but I assume we're pushing through it.

I tried to use our BrowserCam account to verify that it will still work with Safari 1.x. Alas, their Macs don't allow use of ssh, without which I cannot reach my test install. If this patch is deployed in a publicly accessible version somewhere, I would be happy to try it out.

Comment 12

[Gervase Markham [:gerv]]

Patch now deployed on 
https://landfill.bugzilla.org/bug441496/index.cgi

Feel free to test :-)

Gerv

Comment 13

[Guy Pyrzak]

I used the url above. I searched for the word test and nothing came back. At all. And the page appeared to be done loading. I also was unable to view the source. I am using Safari version 3.1.1 (5525.18)

Comment 14

[Bob Williams]

Hmm, same thing here, also with 3.1.1.

Poking my head into the traffic flow, it looks like there's an internal server error being returned without a body. I switched Safari's UA string to a couple of 1.x versions and got the same result. Here's the full request:

Start Watching Built-in Ethernet Traffic, with Filter HTTP

TCP packet from 192.168.0.26:56862 to 63.245.210.140:http(80) (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 64
 Identifier: 55351; Do Not Fragment; Fragment Offset: 0
 Time To Live: 64; Protocol: 6; Header Checksum: 0
TCP Header:
 Sequence Number: 2857218913; Acknowledgment Number: 0
 Header Length: 44; Code Bits: SYN; Window: 65535
 Checksum: 54134; Urgent Pointer: 0
 24 bytes of TCP options:
  Maximum Segment Size: 1460
  Window Scale: 2**1
  Timestamps: 511238997, 0
  SACK Permitted

TCP packet from 63.245.210.140:http(80) to 192.168.0.26:56862 (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 60
 Identifier: 0; Do Not Fragment; Fragment Offset: 0
 Time To Live: 53; Protocol: 6; Header Checksum: 29304
TCP Header:
 Sequence Number: 615939478; Acknowledgment Number: 2857218914
 Header Length: 40; Code Bits: SYN ACK; Window: 5792
 Checksum: 20127; Urgent Pointer: 0
 20 bytes of TCP options:
  Maximum Segment Size: 1460
  SACK Permitted
  Timestamps: 276045503, 511238997
  Window Scale: 2**7

TCP packet from 192.168.0.26:56862 to 63.245.210.140:http(80) (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 61211; Do Not Fragment; Fragment Offset: 0
 Time To Live: 64; Protocol: 6; Header Checksum: 0
TCP Header:
 Sequence Number: 2857218914; Acknowledgment Number: 615939479
 Header Length: 32; Code Bits: ACK; Window: 33304
 Checksum: 54122; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 511238997, 276045503

TCP packet from 192.168.0.26:56862 to 63.245.210.140:http(80) (568 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 620
 Identifier: 58196; Do Not Fragment; Fragment Offset: 0
 Time To Live: 64; Protocol: 6; Header Checksum: 0
TCP Header:
 Sequence Number: 2857218914; Acknowledgment Number: 615939479
 Header Length: 32; Code Bits: ACK PUSH; Window: 33304
 Checksum: 54690; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 511238997, 276045503
Data:
GET«spc»/bug441496/buglist.cgi?query_format=specific&order=relevance+desc&bu
g_status=__open__&product=&content=e«spc»HTTP/1.1«cr»«lf»
User-Agent:«spc»Mozilla/5.0«spc»(Macintosh;«spc»U;«spc»PPC«spc»
Mac«spc»OS«spc»X;«spc»en)«spc»AppleWebKit/125.5.7«spc»(KHTML,«spc»
like«spc»Gecko)«spc»Safari/125.12«cr»«lf»
Cache-Control:«spc»max-age=0«cr»«lf»
Referer:«spc»http://landfill.bugzilla.org/bug441496/query.cgi«cr»«lf»
Accept:«spc»text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,t
ext/plain;q=0.8,image/png,*/*;q=0.5«cr»«lf»
Accept-Language:«spc»en-us«cr»«lf»
Accept-Encoding:«spc»gzip,«spc»deflate«cr»«lf»
Cookie:«spc»DEFAULTFORMAT=specific«cr»«lf»
Connection:«spc»keep-alive«cr»«lf»
Host:«spc»landfill.bugzilla.org«cr»«lf»
«cr»«lf»

TCP packet from 63.245.210.140:http(80) to 192.168.0.26:56862 (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 25460; Do Not Fragment; Fragment Offset: 0
 Time To Live: 53; Protocol: 6; Header Checksum: 3852
TCP Header:
 Sequence Number: 615939479; Acknowledgment Number: 2857219482
 Header Length: 32; Code Bits: ACK; Window: 55
 Checksum: 37240; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 276045539, 511238997

TCP packet from 63.245.210.140:http(80) to 192.168.0.26:56862 (185 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 237
 Identifier: 25461; Do Not Fragment; Fragment Offset: 0
 Time To Live: 53; Protocol: 6; Header Checksum: 3666
TCP Header:
 Sequence Number: 615939479; Acknowledgment Number: 2857219482
 Header Length: 32; Code Bits: ACK PUSH; Window: 55
 Checksum: 18186; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 276045644, 511238997
Data:
HTTP/1.1«spc»500«spc»Internal«spc»Server«spc»Error«cr»«lf»
Date:«spc»Fri,«spc»27«spc»Jun«spc»2008«spc»22:36:49«spc»
GMT«cr»«lf»
Server:«spc»Apache/2.2.3«spc»(CentOS)«cr»«lf»
Content-Length:«spc»0«cr»«lf»
Connection:«spc»close«cr»«lf»
Content-Type:«spc»text/plain;«spc»charset=UTF-8«cr»«lf»
«cr»«lf»

TCP packet from 192.168.0.26:56862 to 63.245.210.140:http(80) (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 9028; Do Not Fragment; Fragment Offset: 0
 Time To Live: 64; Protocol: 6; Header Checksum: 0
TCP Header:
 Sequence Number: 2857219482; Acknowledgment Number: 615939664
 Header Length: 32; Code Bits: ACK; Window: 33211
 Checksum: 54122; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 511238998, 276045644

TCP packet from 63.245.210.140:http(80) to 192.168.0.26:56862 (628 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 680
 Identifier: 25462; Do Not Fragment; Fragment Offset: 0
 Time To Live: 53; Protocol: 6; Header Checksum: 3222
TCP Header:
 Sequence Number: 615939664; Acknowledgment Number: 2857219482
 Header Length: 32; Code Bits: ACK PUSH; Window: 55
 Checksum: 4106; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 276045644, 511238997
Data:
<!DOCTYPE«spc»HTML«spc»PUBLIC«spc»"-//IETF//DTD«spc»HTML«spc»
2.0//EN">«lf»
<html><head>«lf»
<title>500«spc»Internal«spc»Server«spc»Error</title>«lf»
</head><body>«lf»
<h1>Internal«spc»Server«spc»Error</h1>«lf»
<p>The«spc»server«spc»encountered«spc»an«spc»internal«spc»
error«spc»or«lf»
misconfiguration«spc»and«spc»was«spc»unable«spc»to«spc»complete«lf»
your«spc»request.</p>«lf»
<p>Please«spc»contact«spc»the«spc»server«spc»administrator,«lf»
«spc»[e-mail]«spc»and«spc»inform«spc»them«spc»
of«spc»the«spc»time«spc»the«spc»error«spc»occurred,«lf»
and«spc»anything«spc»you«spc»might«spc»have«spc»done«spc»
that«spc»may«spc»have«lf»
caused«spc»the«spc»error.</p>«lf»
<p>More«spc»information«spc»about«spc»this«spc»error«spc»may«spc»
be«spc»available«lf»
in«spc»the«spc»server«spc»error«spc»log.</p>«lf»
<hr>«lf»
<address>Apache/2.2.3«spc»(CentOS)«spc»Server«spc»at«spc»landfill.bugz
illa.org«spc»Port«spc»80</address>«lf»
</body></html>«lf»

TCP packet from 63.245.210.140:http(80) to 192.168.0.26:56862 (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 25463; Do Not Fragment; Fragment Offset: 0
 Time To Live: 53; Protocol: 6; Header Checksum: 3849
TCP Header:
 Sequence Number: 615940292; Acknowledgment Number: 2857219482
 Header Length: 32; Code Bits: FIN ACK; Window: 55
 Checksum: 36321; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 276045644, 511238997

TCP packet from 192.168.0.26:56862 to 63.245.210.140:http(80) (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 9523; Do Not Fragment; Fragment Offset: 0
 Time To Live: 64; Protocol: 6; Header Checksum: 0
TCP Header:
 Sequence Number: 2857219482; Acknowledgment Number: 615940292
 Header Length: 32; Code Bits: ACK; Window: 32990
 Checksum: 54122; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 511238998, 276045644

TCP packet from 192.168.0.26:56862 to 63.245.210.140:http(80) (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 30212; Do Not Fragment; Fragment Offset: 0
 Time To Live: 64; Protocol: 6; Header Checksum: 0
TCP Header:
 Sequence Number: 2857219482; Acknowledgment Number: 615940293
 Header Length: 32; Code Bits: ACK; Window: 32990
 Checksum: 54122; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 511238998, 276045644

TCP packet from 192.168.0.26:56862 to 63.245.210.140:http(80) (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 10297; Do Not Fragment; Fragment Offset: 0
 Time To Live: 64; Protocol: 6; Header Checksum: 0
TCP Header:
 Sequence Number: 2857219482; Acknowledgment Number: 615940293
 Header Length: 32; Code Bits: FIN ACK; Window: 33304
 Checksum: 54122; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 511238998, 276045644

TCP packet from 63.245.210.140:http(80) to 192.168.0.26:56862 (0 bytes)
IP Header:
 Version: 4; Header Length: 20; TOS: 0; Packet Length: 52
 Identifier: 25464; Do Not Fragment; Fragment Offset: 0
 Time To Live: 53; Protocol: 6; Header Checksum: 3848
TCP Header:
 Sequence Number: 615940293; Acknowledgment Number: 2857219483
 Header Length: 32; Code Bits: ACK; Window: 55
 Checksum: 36288; Urgent Pointer: 0
 12 bytes of TCP options:
  Timestamps: 276045675, 511238998

Stop Watching Network Traffic

Comment 15

[Gervase Markham [:gerv]]

Guy, Bob: that wasn't my patch. I backed it out and still saw a problem. I cvs updated and it went away, so please try again :-)

Gerv

Comment 16

[Bob Williams]

Unfortunately, it seems to have returned, as I tried to test and it was happening again. Here are the response headers as reported by Firefox:

Date: Tue, 01 Jul 2008 17:07:47 GMT
Server: Apache/2.2.3 (CentOS)
Content-Length: 0
Connection: close
Content-Type: text/plain; charset=UTF-8

500 Internal Server Error

Comment 17

[Gervase Markham [:gerv]]

Ah, I see. "Editing" or patching buglist.cgi changes its group and then Apache can't read it. Try again now.

Gerv

Comment 18

[Bob Williams]

I tested in Safari 1.3.x, Safari 3.1.1, and OmniWeb 5.7 (which uses a modified version of WebKit), and everything seemed to work as expected. I got the intermediate page followed by the results page in Safari 3 and OW, and went straight to the results page in Safari 1.3.

Comment 19

[Gervase Markham [:gerv]]

Checking in buglist.cgi;
/cvsroot/mozilla/webtools/bugzilla/buglist.cgi,v  <--  buglist.cgi
new revision: 1.377; previous revision: 1.376
done

Gerv

Comment 20

[Frédéric Buclin]

Backing out the patch. It breaks Safari 3.2.3 on Mac and Google Chrome 2.0 on Mac and Windows, see bug 492946.

For the record, the patch never landed on the 3.2 and 3.0 branches despite the target milestone and approval flags, so only 3.3 and 3.5 are affected by this problem.

tip:

Checking in buglist.cgi;
/cvsroot/mozilla/webtools/bugzilla/buglist.cgi,v  <--  buglist.cgi
new revision: 1.397; previous revision: 1.396
done

3.3.4:

Checking in buglist.cgi;
/cvsroot/mozilla/webtools/bugzilla/buglist.cgi,v  <--  buglist.cgi
new revision: 1.394.2.2; previous revision: 1.394.2.1
done

Comment 21

[Dave Miller [:justdave]]

Sounds to me like Safari implemented this in 3.1 and had it working correctly there, then broke it in 3.2 (and other browsers that use the corresponding version of WebKit)

Comment 22

[Gervase Markham [:gerv]]

/me breathes a sigh of relief that it wasn't his fault.

I say we give up on Safari server-push until they get their act together. So I suggest WONTFIX.

Gerv

Comment 23

[Marcus Pallinger]

It seemed to be working well on the first beta of Safari 4.0 on Mac OS X

Comment 24

[Gervase Markham [:gerv]]

WONTFIX - server push on Safari seems to be intermittent. They can cope without :-)

Gerv

Comment 25

[Max Kanat-Alexander]

Agreed. Until such time as all supported versions of Safari reliably support push, we shouldn't push to Safari.