Re: Updated Benchmark Available

[Paul Tchistopolskii <paul at qub dot com>]

----- Original Message ----- 
From: Alexey Gokhberg <alexei@bluewin.ch>

> > ... The topic "is  XSLT fast enough"  is in fact
> > very tricky and requires long discussion ...
> > 
> 
> Sure. 
> 
> But I beleive that this has more to do with the implementation, than
> with the XSLT specification itself. From some point of view, XSLT could
> play for XML the same role as SQL plays for relational data. In
> particular, in both cases the sophisticated optimization techniques are
> needed to achieve the reasonable performance. I am pretty sure, in the
> near future we will see the new generation of optimized XSLT processors.

So am I. But I'm also sure that some of those iplementations will not 
be 100% conformant to XSLT WD.  By the way  what is SQL ? ;-)  
I think "100% portable SQL queries" have not too much sense in 
current world.
 
> Indeed, I am concerned more with the topic "is developing stylesheets
> with XSLT fast enough". As I can understand, XSL was concieved as a
> declarative language that does not require substantial programming
> skills from the stylesheet authors. However, as we could see so far,
> implementation of many relatively simple algorithms may require advanced
> LISP-like methods. Who knows how much man-hours will be spent for doing
> in XSLT things that could be easily done in other languages? and how
> many people will experience difficulties mastering XSLT? Yes, there are
> few XSLT gurus on the Web, they are glad to help everyone, but is their
> summary manpower sufficient to support the industrial use of XSLT across
> the world?

Sure I agree ( because I was constantly making the same statements 
myself ). There is not a big number of people who can get pipes, there is 
not a big number of developers who get functional programming. Some 
people are saying: "this is issue of education". I doubt ( in fact it 
*is* issue of education, but not the issue of 'programming education' )
I mean making functional language to be a 'first language' has no serious 
impact. 7+ years ago I have been involved in some experiments when 
one class of students (children) got  Pascal as a first language 
when the second class got Logo ( I was used to write part of MIT-logo 
interpreter ). I came to the conclusion that ability to use concepts of 
functional programming has nothing to do with the 'programming educaion'.
I think masses will use not more than  3-4 constructions of XSLT. But that's 
not too bad actually.

> Please, understand me correctly. I appreciate XSLT technology very much
> (having invested a lot in it), but I am really afraid that if the
> usability issues will be not successfully resolved, XSLT can die very
> soon, despite those bright ideas that form the basis of this technology.

XSLT will not die ;-) Well ... not really. I consider XSLT to be a first 
prototype implementation of some important concepts. Syntax is weird, 
many useless hacks are in the core e t.c. e t.c. But that's not a big deal.

XSLT contains some brilliant inventions - those inventions will not 
die. For example, I mean the biggest invention of XSLT is that 
easy combination of  'push-pull' is the way to go. Remove pull 
from XSLT and we'l get awk / M4. Remove push - we'l get kinda SQL.

The next steps will take into account the importance of push-pull 
combination - this means XSLT will not 'die'. This is in fact a long topic 
and I'm not sure we should start it now. 

> ... but whether it is smart or just wise, it got the point - it
> attracted much more attention than the ordinary announcement could do.
> This is like a street fight - nobody will judge how refined is your
> technique, only the final result is important.

Hm. Since in the 60-s(?)  in the US they invented that it is more profitable 
to invest  $1 into advertizing of beer rather than invest that $1 into 
quality of beer - thats the way marketing goes. Of course this also explains
why there is no good beer in the US. I think those who care about the quality 
of beer should just ignore those who are concerned about marketing.

Some time ago I've wrote a long letter - prediction about the future of XML.
My prediction is that at some point those of us who care about the quality 
of beer will have to start with 'another XML' - like Linux was 'another UNIX'.

I think that if you just move your stuff in reasonable direction - you can 
be sure - at some point those who are making real job will recognize 
your product no matter how conformat it will be at that point ;-)

Of course, as I wrote in my prediction, this is not a good thing when 
those who care about some XML parts are working 'against' W3C
( like it was not a good thing that Linux started working against 
FreeBSD project ). It will be of course better if, for example, 
you and some other people who care about usability of 
XML-related things will be on WG, or something. 

But this not gonna happen ever, I think. FreeBSD people 
were not smart to recognize the possible impact of 
Linus Torvald. In result we got Linux and FreeBSD when 
we could get one project ( and one developers base ).

But  maybe it was *good* we got FreeBSD / Linux.
Maybe it will be *good* if we get "W3C alternative".

This is the issue of time, I think. How long people will 
keep waiting for XSL FO ? They'll stop waiting when 
they'll *really* start using  XML ;-) 

Rgds.Paul.




 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list