[XSL-LIST Mailing List Archive Home] [By Thread] [By Date]

Re: [xsl] New XSLT 3.0 Working Draft

Subject: Re: [xsl] New XSLT 3.0 Working Draft
From: Michael Kay <mike@xxxxxxxxxxxx>
Date: Thu, 12 Jul 2012 18:37:46 +0100

On 12/07/2012 16:34, Syd Bauman wrote:
I think Tommie says this well. But rather than focus on the
unsupported (and IMHO likely to be untrue) premise, I'd like to focus
for a moment on the question: "what are the chances of 3.0 being

My thoughts on this are

* while for users like me the advancements in 3.0 are way cool, but
   not particularly important,

* for users like Google and Amazon who have *lots* of data, the
   streaming capability of 3.0 will make it very attractive,

* users like Google and Amazon have money and resources, where users
   like me don't,

* it's extremely likely that at least 1 major implementation (Saxon)
will support 3.0,
I think it pretty likely that 3.0 will have widespread adoption as a
programming language and for use server-side. I doubt it will have
any feet in-the-browser.

One thing we haven't really looked at yet is profiling. Before the spec is carved in stone, I would expect us to define some features (such as streaming, packages, maybe higher-order-functions) as optional. That will hopefully leave a spec which is still a useful advance for the mainstream user (with things like maps, try/catch, and evaluate) while not being vastly expensive for an existing XSLT 2.0 implementation to upgrade to.

As for the browser, we've just got the first XSLT 2.0 implementation and we'll have to see how it fares. It's certainly true that client-side usage hasn't been a major influence on 3.0 requirements, though there are a few features like try/catch and evaluate that would be very useful there. I'm a little tempted to "cherry-pick" such features into Saxon-CE, but we'll wait a bit for the user feedback before doing that.

I think you're right to talk about money. This stuff can't be implemented by hobbyists unless they are seriously committed. At the same time, it's competing with free stuff that creams off the bulk of the users (both XSLT 1.0 and alternative technologies). That means the investment comes from people who are getting serious value from the technology, which is why the new features end up being geared towards the needs of the minority of users who have big requirements. Unfortunately these users don't tend to talk much about what they are doing, which can easily create the impression that there is nothing happening at the top end. (I don't even know myself what my biggest users are doing: I only know that they are prepared to pay for it.)

Michael Kay

Current Thread