66858 - [RFE] cached style sheet not used from disk cache

Status: RESOLVED DUPLICATE of bug 29370

(Core :: Networking: Cache, enhancement)

Description

[HJ]

If a site uses linked Style Sheet, they pop up in disk cache. But after this the
style sheets are always, at least for WinNT4, fetched from the netwerk (server).
The correct behavour would be to use the cached style sheets, and not to force a
reload.

Comment 1

[Jure Repinc (JLP)]

Fixed spelling error in Summary (cashed -> cached)

Comment 2

[Mirek Hankus]

I think it is dup of #29370

Comment 3

[HJ]

Thank you, high five. And thank you Mirek, but this bug is about disk cache.
Reading the bug writer guidelines, please use only one issue per bug report.

Also in bug report # 66803 Gordon told me: "The memory cache is currently only
used by IMAP". But maybe someone took out Memory support to 'hack around' some
of his problems!?

And do have any idea what the performance gain is if you don't have to reload
_all_ linked Style Sheets? Well I do, but then, how many people do you think are
using a protocol analyzer? 

The fact that someone is inserting cache header information brakes HTTP 1.1 (RFC
2116) compliance, and that's seems very important to me! Something I did not see
in that other bug report!

Comment 4

[gordon]

cc'ing darin because of possible implications for HTTP.

Comment 5

[Darin Fisher]

reporter: can you verify that the cached CSS items have not expired?  thx!

Comment 6

[HJ]

No timeout, and I did see a line with something like FORCERELOAD in the source,
that won't help I guess!?

Comment 7

[HJ]

Ok cheched, Prage and Cache-control are set to nocache. And in fact if a
document has expired it should perform a GET not-changed-since request and the
server should reply with a 304, not modified response, right?

Comment 8

[Darin Fisher]

marking bug INVALID since cached HTTP headers indicate that server expects
client to do a fresh GET on each request for the URL.

Comment 9

[HJ]

Darin tell me something, this counts for Mozilla only, why? Microsoft Internet
Explorer, Opera and Amaya don't have this in the header. It's the same server,
it are the same documents. They all use HTTP 1.1 so what's the difference then?

The first GET don't have Pragma and Cache-Control set to nocache. So who put
that in the header? And this seems to count for ALL servers, for ALL Javascript
and ALL CSS files! But this doesn't count for HTML and images!

I really don't buy this, this is clearly a bug in Mozilla because it's clearly a
Mozilla bug.

Comment 10

[HJ]

This is interesting:

nsCSSLoader.cpp
1224 if (isLocalFile)
1225 loadAttributes = (nsLoadFlags)nsIChannel::LOAD_NORMAL;
1226 else
1227 loadAttributes = (nsLoadFlags)nsIChannel::FORCE_RELOAD;
1228 result = NS_NewStreamLoader(&loader, urlClone, aData, nsnull, loadGroup,
1229 nsnull, loadAttributes);

The ball is yours.

Comment 11

[Darin Fisher]

Anyone know why we FORCE_RELOAD online CSS content?

Comment 12

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

CVS log for nsCSSLoader.cpp, rev. 3.36:

3.36 pierre%netscape.com Mar 8 2000   29370 "Stylesheets and JS files are not
always loaded". Checked in for valeski, r=pierre, a=jar.

see bug 29370

Comment 13

[Darin Fisher]

Ok, looks like valeski's patch for bug 29370 was a work around because of
problems with the necko cache.  Reopening, so we can fix this when the new
cache lands.

Comment 14

[Darin Fisher]

over to gordon

Comment 15

[Judson Valeski]

indeed a workaround.

Comment 16

[HJ]

Has network speed to do with this path? If so, can this work on a SAT connection
without the path? I use an ASTRA VPN SAT connection, where I can use Mozilla on.
The average speed on this SAT is about 8Gb/sec per sec, but I can easily switch
it up to 19 Gb/sec, as I do for downloads. This SAT link is for Europe only, but
we do have an international SAT link also, up to 38Gb/sec. I use this link for
video conferences with the US and I receive file updates via this link.

Comment 17

[HJ]

BTW, "HTTP 1.1 (RFC 2116)" should read "HTTP 1.1 (RFC 2616)" typo.

Comment 18

[Tim Powell]

Shouldn't this be marked as a duplicate of bug 29370 or at least changed from
RFE/enhancement? This severely impacts the usefulness of CSS. Taking a reloading
hit each time is a bug.

Comment 19

[Darin Fisher]

Agreed.  Marking as a dup of bug 29370.

*** This bug has been marked as a duplicate of 29370 ***

Comment 20

[Hixie (not reading bugmail)]

Darin: Are you sure? That other bug is for the memory cache, no?

Comment 21

[Darin Fisher]

The summary of bug 29370 should probably be updated.  Jud's "temporary" patch
from that bug causes us to skip all caching of CSS; not just caching in memory.

We only need to track this bug in one place, so I picked bug 29370.