Re: [SVGMobile12] Too many issues

On Fri, 20 May 2005, Jon Ferraiolo wrote:
> While clearly these sections have shortcomings and indeed should be 
> improved, and thanks for all of your comments, I think it is worth 
> pointing out that: (a) somehow these sections managed to get through 
> multiple Last Call, Candidate Recommendation, Proposed Recommendation 
> and Recommendation milestones in the past

I can't speak for the reviewers of those previous versions; I didn't spend 
much time looking at the SVG specs in the past, for various reasons. Most 
of the issues I raised back in the SVG 1.0 time frame were ignored and are 
still not resolved today (e.g. unitless values, which have caused 
headaches ever since to those of trying to implement SVG and CSS 
together), so I'm not convinced that just because the text made it past 
last call that means that the text is appropriate phrased.

In any case, many of the problems that I've raised so far seem to be 
related to the fact that the introductory chapters were written for a spec 
with different features and different conformance requirements. Thus, 
while the text may have been appropriate for SVG 1.1, it does not mean it 
is appropriate for SVG 1.2.

> There is plenty of evidence in the form of dozens of commercial 
> implementations that the introductory sections which have generated your 
> 25 issues have been consistent with the notion of implementability.

In all fairness, I would say the exact opposite. As you point out, Opera 
has been working on an SVG implemetation. As part of that we have had to 
make our own tests to check interoperability, and have been testing 
interoperability between our products and other vendors'. What we have 
found is that SVG has huge interoperability problems, is very 
inconsistently implemented, and that the publicly available test suites 
are inadequate at best.

For instance, we found _no_ implementations that implemented the 
conformance criteria in SVG 1.1 (error handling section). We also found 
dozens of implementations that failed utterly to correctly support 
namespaces (including products from Adobe) and hundreds of images that 
were non-conformant in fundamental ways (e.g., again, missing namespaces).

It is also clear from the number of issues that the Mozilla team are 
running into when implementing SVG (each of those issues being raised on 
this mailing list) that it is not currently possible to derive an 
implementation unambiguously from the existing specifications.

If your experience differs from this I would respectively suggest that 
your testing is inadequate.

Ian Hickson               U+1047E                )\._.,--....,'``.    fL       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Friday, 20 May 2005 19:00:30 UTC