Re: "image analysis heuristics" (ISSUE-66)

On Sun, 7 Feb 2010, Steven Faulkner wrote:
> 
> As a general rule, people don't follow references.
> 
> Can you provide support for this statement?

Sure. Over the years I've worked for Netscape and Opera, as well as 
contributing to the Mozilla, WebKit, and Chromium projects, and regularly 
advising Microsoft. In all of these cases, I have repeatedly seen 
competent engineers do the minimum amount of reading possible to implement 
the feature that they have been tasked to implement. Again in all cases, I 
have found them to implement things better when they are faced with clear 
unambiguous steps to implement, rather than when they are presented with 
constraints; and (more germane here) I've found them to treat suggestions 
inlined in a document with a _lot_ more weight than suggestions found in 
documents referenced from the document they are reading.

I intend to make no value judgements here, I'm merely describing what I've 
found to be true, repeatedly, over the years.

I have no data to support this, and would in no way suggest that my 
experience is The Truth, or try to enforce my conclusions on other 
editors. However, I _do_ intend to take the above into account when 
writing the specs that I edit; I consider maximising the extent to which 
the document is an effective tool for getting quality implementations to 
be part of the responsibility of writing a spec.


> it would seem to me that a link to a reference is like the many 
> thousands of links that the spec already contains, it may be that the 
> phrasing of the link text can affect the likelyhood of a person to 
> follow the link.

As far as I am aware, the only links from the WHATWG complete.html spec to 
other specs are links to specs that are required to be implemented because 
they form the substrate on which HTML and its APIs are built, such that 
the implementors _cannot_ skip them even if they are tempted to.


> This infomation is meant for browser developers is it not?

Yes.


> If they are not interested in making their browsers provide more 
> accessible content, it does not matter how much content you put in the 
> html5 spec, they can easily skip over it.

In practice in my experience most implementors are in principle in favour 
of making their implementations accessible, but on average they are more 
likely to do the right thing when they find the information right there in 
the prose they are having to read anyway, than if it is "conveniently out 
of sight".

As the adage goes, "out of sight, out of mind".


> If they ar interested it would be better, i think, to point them to a 
> document that provides comprehensive advice on how to do so.

We do provide a link to UAAG. However, if there is advice in the UAAG spec 
that you think implementors should follow here, then the best way we can 
ensure that it is followed is, IMHO, to also include it in HTML. Is there 
something I've omitted that UAAG recommends of relevance here?

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

Received on Sunday, 7 February 2010 10:07:29 UTC