Re: Translation status pages

Today at 9:29, Lucas Vieites wrote:

>   Creation of "hundreds of static .html files" seems like a prehistoric
> thing to do. Would it be too complicated to create a few xml files with
> the statistics data (or maybe insert it in a database) and then generate
> the html dinamically (to avoid too much intensive db access, these could
> be cached upon generation to avoid duplication). The flow would be like
> this:

Please read my mail completely.  We are using a pretty old code base,
and new code base which is in the works for some time itself already
does it that way.  We just need someone to finish the new code.

As for "caching upon generation", it's still the pretty much same
thing, you just delay it for a later moment.

>   -process generates data (xml or sql)
>   -user accesses a page first time (eg. /gnome-2.12/es/desktop/)
>     -php generates the page from data and stores it in a temp dir
>   -other user visits the same page
>     -it gets shown from the temp dir (no duplicate generation, no db
> overhead)
>   -temp dir gets erased every time the process runs

HTML generation is still a very minor part of the process.  I just
mentioned it for the sake of disk and filesystem cache invalidation
occuring due to buffering of a large number of files which might be
put all around the disk (though, this is itself unlikely).

(PHP is not a very good choice because of limited security if run as a
module: we don't want another breakage; SuExec solutions are
slower and more demanding, but IMHO worth the costâof course, you can
run PHP through SuExec as well).

>   Questions:
>   -Pages that are never visited never get generated, is this a pro or a
> con?

All pages will likely be visited at least by a webbot.  Anyway, those
not visited are not needed.

>   -Is this really less cpu intensive? I don't know if the sum of cpu
> time for all the pages that are visited (for the first time) is less
> than what we have now.

No, this is more CPU intensive.  But, generating HTML is not about CPU
intensive (either dynamically from a database or by generating static
HTML files) but about using disk (invalidating filesystem and disk
caches, etc.), where this might turn out to be an improvement.  But
I can't tell if it would help or not, since it depends on the usage
patterns of, which I don't have :)

>   Anyway, it's just an idea, I would volunteer to help with this if
> possible.

As I said, try picking up Carlos' new code from somewhere on  He might be able to give a hand himself if you need
it :)


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]