- From: Robert Burns <rob@robburns.com>
- Date: Wed, 22 Aug 2007 16:24:50 -0500
- To: James Graham <jg307@cam.ac.uk>
- Cc: Sander Tekelenburg <st@isoc.nl>, public-html@w3.org
Hi James, On Aug 22, 2007, at 3:56 PM, James Graham wrote: > > Sander Tekelenburg wrote: >>> Henri gave some good examples of problems people try to solve that >>> aren't real in IRC: >>> >>> <hsivonen> Regarding fantasy problems: the thing is people do try to >>> solve them. sometimes the problems aren't 100% fantasy, but >>> common sense >>> says they aren't real problems. Examples: 1) "search engines could" >> I really don't understand how this is a good example. "search >> engines could" >> sounds like relevant context was omitted. > > As I understand it, Henri's point is that it is not uncommon to see > arguments like "We should add elements <x>, <y> and <z> with > attributes a,b and c so that search engines can use the information > they provide", without ever asking people who have experience > developing search engines whether <x>,<y> and <z> or a,b and c > would actually help. Often it turns out that they would not > because, for example, the features only function as intended if > they are use correctly by a large proportion of the userbase, > something that is unlikely to occur in practice (e.g. [1]). > > With this in mind, we should not assume problems are real until we > have spoken to the constituents who are supposedly having the > problem to ask them if it is, in fact, a problem they are having, > and whether the proposed solution will actually have any bearing on > the problem. However, the problem with this line of thinking (and dismissing this as not a real problem), is that this creates a spiraling race to the bottom. If the recommendations to not define clearly how authors should use "elements <x>, <y> and <z> with attributes a,b and c ", then authors won't use them consistently and in large numbers. If authors do not use them consistently and in large numbers, then there's no reason for a search engine to make use of those semantics in indexing content. Then to make the spiral a complete absurdity, we want to codify this by saying that if these features do not exist and therefore authors are not using then and therefor UAs make no use of them, then we should not add them. Therefore, it is not at all the case that these are not "real problems". Instead it is the case that these are real problems that we do not want to solve. Or they are real problems that we do not want to solve by introducing "elements <x>, <y> and <z> with attributes a,b and c". Either way, it's a design principle that I would never support and, frankly, makes no sense to me whatsoever. The other issues raised by Henri in that IRC discussion[1] are quite similar. They are all examples of attempts to solve real problems. They may be problems that already have solutions that are available but not widely used. Or they are problems that Henri felt not worth solving. In that latter sense it is an excellent example of how this design principle is really only designed for mis-use. [1]: <http://krijnhoetmer.nl/irc-logs/html-wg/20070818#l-45>
Received on Wednesday, 22 August 2007 21:25:06 UTC