On Mon, Sep 27, 2010 at 11:13 AM, Martin Scharrer <[email protected]> wrote: >> >>>> I've then started to implement a cache for these things, so that the >>>> rendering happens only once, but got stuck somewhere in the middle. If I >>>> remember correctly, the latest state of that work is not yet available >>>> in the repository. Anyway, Trac meanwhile implements its own cache >>>> implementation, which is much better than what I came up with and which >>>> thus should be used for our purposes. I have to look at that stuff >>>> again. >>> >>> So, Trac 0.11 doesn't have this cache, so we won't get any improvement >>> there, but why on 0.10 this works faster?? >> >> Well, I think in the end the speed issues are caused by Genshi. It's a >> known fact that Genshi is generally a bit slower than Clearsilver. In >> addition, there have been a lot of changes in how the hacks listings are >> generated (just compare the hacks list on both, the live and the testbed >> site). > Maybe the homepage should not list all hacks then but link to [a] extra > page(s)? Caching the list and especially the descriptions, at least > somehow, would be essential anyway.
Compile page into static HTML and regenerate it when reflected information is changed. Just need a list of events when this should be done. Like: 1. added/removed tags 2. added/removed hack Anything else? > Yeah, Debian's two versions: either unstable or outdated ... > > I agree, it doesn't have to be "bleeding", but I ran in the trouble > under my Ubuntu installations (which is Debian based) where I had to > manual compile it to get 1.6. It is a major upgrade from upstream, so it > should be supported. I went from Gentoo to Ubuntu to avoid compiling. :-) > > Another example was this thing with gvim which spamed the console with > hundreds of trivial messages at start-up where they refused to > "backport" the also trivial fix, released just in the week after the > official release. All in the holy name of "stability". These versions are "stable" only when users report the issues, but Debian tracker suxx so greatly that I don't report anything anymore (unless the server is down or hacked). If nobody reports - there is a feeling that everything is "ok", but that's not always true. Yes, this at least keeps flaky Linux package/binary-dependency system user-friendly, but this greatly outdated. I doubt there a could be a policy of "common sense" to solve this Debian problem, so we have to live with it until there will be a better OS, or social-oriented tools, where users can actually influence backporting and upgrading process. It could be Ubuntu, but GPL suxx, so not many professionals want to make their hands dirty with code they won't be able to copy/paste in some future project even with due attributions. -- anatoly t. _______________________________________________ th-users mailing list [email protected] https://lists.trac-hacks.org/mailman/listinfo/th-users
