Bug 226083

Summary:

Don't hang onto expired resources without validation headers in memory cache

Product:

WebKit

Reporter:

Antti Koivisto <koivisto>

Component:

Page Loading

Assignee:

Nobody <webkit-unassigned>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

beidson, cdumez, ews-watchlist, ggaren, japhet, nham

Priority:

Keywords:

InRadar

Version:

WebKit Nightly Build

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
patch	none
patch	ews-feeder: commit-queue-
patch	ews-feeder: commit-queue-
patch	none
patch	none
patch	ews-feeder: commit-queue-
patch	none
patch	none
patch	none

Antti Koivisto

Reported 2021-05-21 07:23:54 PDT

They use memory while only being useful for history navigation. Disk cache can handle that.

Attachments
patch (2.21 KB, patch) 2021-05-21 08:14 PDT, Antti Koivisto	no flags	Details Formatted Diff Diff
patch (2.21 KB, patch) 2021-05-21 08:18 PDT, Antti Koivisto	ews-feeder: commit-queue-	Details Formatted Diff Diff
patch (8.33 KB, patch) 2021-05-24 12:13 PDT, Antti Koivisto	ews-feeder: commit-queue-	Details Formatted Diff Diff
patch (8.41 KB, patch) 2021-05-25 05:19 PDT, Antti Koivisto	no flags	Details Formatted Diff Diff
patch (9.46 KB, patch) 2021-05-25 06:20 PDT, Antti Koivisto	no flags	Details Formatted Diff Diff
patch (8.77 KB, patch) 2021-05-25 07:36 PDT, Antti Koivisto	ews-feeder: commit-queue-	Details Formatted Diff Diff
patch (10.87 KB, patch) 2021-05-26 01:51 PDT, Antti Koivisto	no flags	Details Formatted Diff Diff
patch (10.29 KB, patch) 2021-05-26 01:53 PDT, Antti Koivisto	no flags	Details Formatted Diff Diff
patch (10.63 KB, patch) 2021-05-26 07:59 PDT, Antti Koivisto	no flags	Details Formatted Diff Diff
Show Obsolete (8) View All Add attachment proposed patch, testcase, etc.

Antti Koivisto

Comment 1 2021-05-21 07:44:42 PDT

rdar://78035612

Antti Koivisto

Comment 2 2021-05-21 08:14:44 PDT

Created attachment 429292 [details] patch

Antti Koivisto

Comment 3 2021-05-21 08:18:35 PDT

Created attachment 429293 [details] patch

Antti Koivisto

Comment 4 2021-05-24 12:13:10 PDT

Created attachment 429552 [details] patch

Antti Koivisto

Comment 5 2021-05-25 05:19:02 PDT

Created attachment 429647 [details] patch

Antti Koivisto

Comment 6 2021-05-25 06:20:06 PDT

Created attachment 429651 [details] patch

Antti Koivisto

Comment 7 2021-05-25 07:36:26 PDT

Created attachment 429653 [details] patch

Antti Koivisto

Comment 8 2021-05-26 01:51:26 PDT

Created attachment 429737 [details] patch

Antti Koivisto

Comment 9 2021-05-26 01:53:19 PDT

Created attachment 429738 [details] patch

Antti Koivisto

Comment 10 2021-05-26 07:59:18 PDT

Created attachment 429755 [details] patch

Chris Dumez

Comment 11 2021-05-26 11:09:49 PDT

Comment on attachment 429755 [details] patch r=me

Geoffrey Garen

Comment 12 2021-05-26 11:11:25 PDT

Comment on attachment 429755 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=429755&action=review > Source/WebCore/ChangeLog:29 > + Remove expired resources without validation headers from the cache if the resource is otherwise deletable. I guess this is the more conservative approach, so maybe it's best. But I wonder if there's any value to expired resources _with_ validation headers in the memory cache? In that case, we're going to do a network round trip to validate anyway, so maybe an IPC round trip to the network process is also fine. Long ago, Dave Hyatt told me that the real win of the memory cache was when it could respond synchronously in the current RunLoop iteration.

Antti Koivisto

Comment 13 2021-05-26 11:51:58 PDT

Reusing a resource from memory cache may also save decoding for images, parsing for stylesheets, text decoding etc. (though decoded versions are thrown out pretty fast anyway for unreferenced resources). But yeah, it might be time to throw out the entire revalidation logic in memory cache. Would certainly simplify things. Part of its job used to be to cover things networking layer wasn't handling well.

Geoffrey Garen

Comment 14 2021-05-26 12:14:06 PDT

> But yeah, it might be time to throw out the entire revalidation logic in > memory cache. Would certainly simplify things. Part of its job used to be to > cover things networking layer wasn't handling well. I guess that's a topic for another bug. But might be worth trying out at some point.

EWS

Comment 15 2021-05-26 12:19:36 PDT

Committed r278119 (238169@main): <https://commits.webkit.org/238169@main> All reviewed patches have been landed. Closing bug and clearing flags on attachment 429755 [details].

Note You need to log in before you can comment on or make changes to this bug.