On Thu, Nov 28, 2013 at 11:56 AM, David Rajchenbach-Teller < dtel...@mozilla.com> wrote:
> As many of you know, Session Restore is something of a performance hog, > for many reasons – we have reports of. One of the reasons is that we > store so very many things in sessionstore.js and sometimes keep stuff > for a very long time. > > As part of bug 943352 & followup, we are considering automatically > cleanup some of the contents of sessionstore.js. Since people have > mentioned "webcompat" and "userdata loss" in the context of > sessionstore.js, I'd appreciate some feedback before we proceed. > > So, here are a few things that I believe we could cleanup: > 1. get rid of closed windows after a while; > 2. get rid of closed tabs after a while; > 3. get rid of old history entries of open tabs after a while; > 4. get rid of POST data of history entries after a while; > 5. get rid of DOM storage contents of open tabs after a while; > 6. get rid of form data content of open tabs after a while; > This would all be tackled after we did other things like getting rid of all history entries for iframes, which won't be restored in any case, right? I'm asking because all of these seem problematic to me, to be honest. The first two the least - I think it's mostly ok to have to go through the history itself to look up URLs of windows/tabs that were closed in a previous session. But for the others, I'd very much expect issues. Right now, our session storage is wonderfully robust (as long as no bugs are encountered) in that I can do things like starting to fill in forms without having to worry about the browser crashing or me forgetting to actually send the form before restarting the browser. As a concrete example, I start writing my Status Board[1] entries during the week, and only send them off on Mondays. If we were to get rid of form data after some period of time, I couldn't do this anymore. Even worse: I might not know when exactly form data is discarded, so it'd *seem* to work just fine for a while, and I might invest quite some time in writing my update, only to lose it all of a sudden because Firefox decided that I don't need this data anymore. Similar scenarios can probably be thought up/occur for 4 and 5, too. As for 3, we could maybe gradually get rid of entries, oldest first. So it wouldn't be a hard cut-off, but a gradual loss of entries which are less and less likely to be of interest, anyway. [1]: http://benjamin.smedbergs.us/weekly-updates.fcgi/ > ... > > Note that we don't have space usage number for each of these (bug 942340 > should provide more insight). > > If anybody feels that we are going to break one million websites (or one > million profiles), we would be interested to hear about this. > > Cheers, > David > > -- > David Rajchenbach-Teller, PhD > Performance Team, Mozilla > _______________________________________________ > dev-platform mailing list > dev-platform@lists.mozilla.org > https://lists.mozilla.org/listinfo/dev-platform > _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
