Re: Mandating respect of some heuristics?

The proposal made by Dominique would go a long way towards addressing the grievances of the mobile developers' community, which is seeing finely tailored and mobile-optimized applications being degraded, rendered less usable, or even inoperative because of a number of transcoder deployments. Indeed, "the approach of favoring
mobile-content creators by default would be a better one for the mobile web ecosystem as a whole."

Concerning specific issues raised by Dominique:

1. Fundamental principle.

"I think the high-level question behind this practical proposal is whether a mobile CT proxy should be allowed to transform content that was developed with mobile in mind or not."

The CTG has been developed in the context of adapting desktop content for mobile devices -- hence there is little rationale to allow transformation of mobile content itself. Mobile developers take care to analyze HTTP request headers and rely upon rich device capability databases (WURFL, DetectRight, Volantis, UAProf...) to deliver pages of an appropriate size, in the correct format, with a suitable character encoding, with pictures of the right dimensions, carefully formatted with special style sheets, etc. There is practically nothing CT-proxies can do to improve upon this -- especially since they do not know anything about application semantics, and they basically do not know more about the end-user devices than application servers.

2. Interpretation of heuristics.

It is very correct that heuristics based on the URL patterns are just that -- heuristics; except for some of those dealing with the domain name as such (but not the URL path), one cannot conclusively infer that they identify mobile sites. 

On the other hand, all MIME types dealing with WAP1 or WAP2 content, or legacy Openwave formats, conclusively and unambiguously identify mobile-optimized content. That includes application/vnd.wap.xhtml+xml (nobody targets XHTML mobile profile at desktops) and all other WAP types. This is not the case for text/html (can be desktop, can be i-Mode), or formally application/xhtml+xml (can be XHTML 1.0/1.1, can be XHTML basic).

In the case of DOCTYPES, the situation is even clearer: all those currently listed under section E conclusively identify mobile-only content. The same applies to <link rel="alternate" media="handheld" href=""/>.

I am not sure about the status of mobileOK and its establishment as a W3C recommendation, but this is also a strong case to take into account. 

I have also suggest considering <link rel="alternate" media="all" href=""/> as being mobile-compatible.

3. Allowable operations.

"A CT proxy shouldn't apply any non-trivial transformation (with an explicit list of these; e.g. only compressing/removing extraneous white space would be allowed)."

One might consider allowing "tidying" -- in the sense of correcting violations of well-formedness. And one MUST allow the transformations explicitly specified by standards. I mentioned two applying to WML: encoding to WBXML and converting WML 1.x to WML 2.0. 

The important point to realize is that once the enclosing markup is protected from transformations, then its embedded objects should be off limits too (images, scripts, style sheets, etc).


So far, the major criticism of transcoders has not been what they perform on desktop content, but what they do  with mobile content. Mandating the heuristics following Dominique's proposal would probably solve a large fraction of the issues raised by the mobile community.

E.Casais

Received on Monday, 1 December 2008 05:22:44 UTC