On Mon, 28 Jul 2003, Aleksandr Guidrevitch wrote: > May be I'm a bit late here... But is there any sence in artifical XML > templating languages since there is XSLT ? Just wonder whether there are > cons other than long learning curve and performance issues ?
Well, in the case of just TAL/Petal, the point as far as I know isn't so much that it happens to be valid XML -- though of course that's a welcome design aspect -- but that it's easy to use in W3C compliant HTML without masking the "interesting bits" inside HTML comment blocks. I can't think of an example now (and apologize for only having had time to skim much of this long, spiralling thread so far), but I seem to remember that TAL's syntax allowed for these clever little "inside out" loop structures that, among other things, ensures that the unrolled version of the loop still has proper indentation etc. Unessential maybe, but a nice touch. I don't know nearly enough about XSLT to comment on it in depth, but my impression is that TAL generally derives from the same heritage as all the assorted Perl template languages, etc, rather than ...well I guess I don't really know where XSLT derives from (Docbook?), but it seems more like CSS: a separate document describing how to display one or more other structured documents. The difference in the case of TAL / TT / H::T / Mason / etc is that the template is used to provide that structure itself, assembling it out of code that is maybe extracting content from a database or calculating it on the fly something. My hunch is that the way to work XSLT into the mix -- if it has a place at all -- is to have your logic (scripts with H::T, code sprinkled in Mason templates, TT applied where clever TT people apply it) fleshing out the framework defined in a template (the H::T template, the non-code part of a Mason template, the plaintext part of a TT template) as well formed XML, and then using XSLT to format that intermediate form for display. This probably makes sense in cases where that intermediate form can be cached (pages for web or print that change once a day, say), but for truly dynamic documents the XSLT stage is probably way too much overhead. Is XSLT appropriate for applying both structure & styling to the free- form output of a [Perl] program? Does it take a lot of overhead? Would this overhead happen on the server end (glue XML + XSLT into an HTML document that goes out over the wire), or on the client end (download XML, then grab XSLT by reference (like HTML grabs CSS), and then the client merges them)? If the client does the work, how widespread is support? If the server does the work, what should the throughput be like? My hunch -- which I'd be happy to be corrected about -- is that this can't be any easier than just working directly from traditional template kits, as has been discussed at remarkable length in this thread :) -- Chris Devers [EMAIL PROTECTED] http://devers.homeip.net:8080/resume/ sorry for helping along in this thread 10 days & dozens of messages ago :)