Closed Bug 648784 (AddonSlowStartup) Opened 13 years ago Closed 12 years ago

Addon Slow Startup tests tracking bug

Categories

(addons.mozilla.org Graveyard :: Public Pages, defect)

defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: nmaier, Unassigned)

References

Details

(Keywords: meta)

Get an overview about what is currently wrong with the Addon Slow Startup tests
Depends on: 647561
Depends on: 648113
Keywords: meta
Depends on: 648793
Is there a bug for adding items to the best practices list? (https://developer.mozilla.org/en/Extensions/Performance_best_practices_in_extensions)

Would be good to add the suggestion about trying varying XPI compression levels, per: 
http://adblockplus.org/blog/some-more-details-on-mozilla-s-add-on-performance-measurements
(which in the case of ABP made a 15% difference to the addon load time)
Blocks: 599169
Yeah I thought as much, I just don't have an MDN login.

Thanks for adding that though :-)
Depends on: 648978
Depends on: 649286
(In reply to comment #2)
> Generally, the Talk Page is a good place for that. Anyway, I added a section:
> https://developer.mozilla.org/en/Extensions/Performance_best_practices_in_extensions#Use_the_right_compression_level_for_JAR_and_XPI_files

Actually, the last sentence in the first paragraph you added about JARs in XPIs is inaccurate, though not wrong. Using 0 compression (i.e. store instead of deflate) for a JAR inside an XPI will not increase the end XPI file size, it will actually _decrease_ it. It sounds counter-intuitive, but this creates a poor-man's solid archive effect allowing compression between files inside the JAR as apposed to just within files in the XPI. This effect also means that the new trend to ditch the JAR in favor of just an XPI actually increases the XPI file sizes, especially in the case of addons with a large number of locale files that could be better compressed with an internal JAR. I've edited the section slightly to mention this.
(In reply to comment #2)
> Generally, the Talk Page is a good place for that. Anyway, I added a section:
> https://developer.mozilla.org/en/Extensions/Performance_best_practices_in_extensions#Use_the_right_compression_level_for_JAR_and_XPI_files

Actually, the last sentence in the first paragraph you added about JARs in XPIs is inaccurate, though not wrong. Using 0 compression (i.e. store instead of deflate) for a JAR inside an XPI will not increase the end XPI file size, it will actually _decrease_ it. It sounds counter-intuitive, but this creates a poor-man's solid archive effect allowing compression between files inside the JAR as apposed to just within files in the XPI. This effect also means that the new trend to ditch the JAR in favor of just an XPI actually increases the XPI file sizes, especially in the case of addons with a large number of locale files that could be better compressed with an internal JAR. I've edited the section slightly to mention this.
Ack, Bugzilla had a server error and said to try again later, but seems to have posted anyway. Sorry for the extra bugspam. :/
I know that it decreases XPI size - but that's really irrelevant in that context.
Depends on: 650778
Depends on: 650790
Depends on: 650947
Depends on: 658187
The automated perf testing has been on hold for a bit but we're ready to start looking at it again.  I'm working with a new team on what we want to accomplish.  The latest plan is to still alert end users of performance problems but only with manually tested add-ons by someone at Mozilla.  A draft of the plan is at https://wiki.mozilla.org/User:Clouserw/AMO/Perf and I'll be finishing that up this week.

I'm going to wontfix this bug and its friends since they apply to the poor data/policies of v1.  If you've got feedback on the new plan let me know, email the list, talk in #amo, etc.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → WONTFIX
Also should mention - thanks to the folks who filed all these bugs.  I know a lot of research/work went into them.
Product: addons.mozilla.org → addons.mozilla.org Graveyard
You need to log in before you can comment on or make changes to this bug.