SUMMARY: alfs direction

pak_lfs at pak_lfs at
Sat Nov 26 05:10:18 PST 2005

Joachim Beckers wrote:
> The only way I think xml would be useful is in a system where the
> dtd/schema would have *every* possible command in it (even the obscure
> ones like bc or so). It wouldn't be to difficult to auto-generate an xml
> profile then. It's just a matter of having the xslt processor extract
> tags from the book and put them in a "profile" in the right order.
> Someone said ant works this way, but I don't have any experience with
> it. Don't know how it works out. But as I said before, I favour ditching
> xml support entirely.
> I hope to see a new tool shortly, so c'mon let's code! ;-)
> Joachim

Unfortunately, even that wouldn't do since you would also need to support
*the command-line syntax of each tool*. For one, this is beyond the expressive
power of XML (at least if you are not willing to jump through MAJOR hoops).

That's because XML is just syntax  and does not care about the contents of
those tags/attributes. So, there is no (simple) way to say that a <sed_param>
tag 's CDATA should be a regexp. So you would need a  DTD + a specification
for tag semantics and that's where things start to go really wrong.

So, again: XML for anything below the profile/stage description / dependency 
resolution layer is a big no-no.


____________________________________________________________________ - äùñåÜí õðçñåóßá çëåêôñïíéêïý ôá÷õäñïìåßïõ. - free email service for the Greek-speaking.

More information about the alfs-discuss mailing list