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

Re: (Possible) menu code rewrite

Henrique de Moraes Holschuh wrote:

> On Tue, 16 Jul 2002, Craig Dickson wrote:
> > Henrique de Moraes Holschuh wrote:
> > > The snippets would need to be all read, and XSLT-processed in the user's
> > > system, every time an entry is updated.
> > 
> > Which only occurs when packages are installed, updated, or removed,
> > right? Even on unstable machines (aside from package developers), that's
> No.  The menu system allows local user overrides, and a lot of other very
> nice and very neat features that make such updates more frequent, I think.

How frequent? Every time you pop open the menu? Only when you change
some user preference in a config dialog? Every time you login?

> I am not sure it is always handled by update-menus, or if some of the menu
> code for window managers actually does dynamic refresh...

I can see how a window manager might re-read the menu every time you pop
it open, just to be sure it's current, but the XSLT transformation from
the source to the window manager's own format should only need to be
done when something actually changes.

> > Well, it should be way more flexible, and way easier for new developers
> > to learn, since it will be based on industry standard tools rather than
> > some homebrew parser/translator.
> Well, complexity is overrated. The menu syntax is very clean and very easy
> to understand. I don't know about the .desktop syntax, though.

Having to write your own parser is more work than not having to write
your own parser by using industry-standard tools.

> > I've been doing a lot of work lately with XML and XSLT, and I haven't
> > found xsltproc to be a performance problem at all. Though in the
> > interest of full disclosure, I should admit that I haven't run it on
> > anything slower than a 700 MHz Celeron with 384MB RAM.
> Yes. And you can bet some of us still want Debian to non-suck on a 486 with
> 16MB of RAM...  If the XSLT transforms are rare, that's not a problem. But
> if they are not...

I have a Pentium-90 at home that I should upgrade to Woody. Once I've
done that, I can try the current xsltproc and see how painful the
performance is for the material I'm working with.

This question of how often the XSLT transform occurs is critical to this
discussion, so we really ought to get a clear sense of that, and also of
XSLT's performance, before jumping to conclusions one way or another
about its acceptability for this purpose.


Attachment: pgp0c2s_k3nnp.pgp
Description: PGP signature

Reply to: