- From: Philip Taylor <P.Taylor@Rhul.Ac.Uk>
- Date: Wed, 27 Nov 2013 19:40:54 +0000
- To: "Jukka K. Korpela" <jkorpela@cs.tut.fi>
- CC: "'www-html@w3.org'" <www-html@w3.org>
Jukka K. Korpela wrote: > 2013-11-27 20:58, Philip Taylor wrote: >> In what way have they been misled, Jukka ? > > The way that they regard a syntax declaration as specifying a version in > a manner that would matter anything outside formal validation. I'm not convinced that people have been "misled"; some may have (mistakenly) assumed that the chosen syntax declaration would matter outside of formal validation, and many more would assert that the chosen syntax declaration /should/ matter outside of formal validation -- I am not convinced that either have been misled. > By the SGML standard, which is normatively the only thing that matters > here, you are just specifying a DTD, a formal syntax specification. > Validators will by default use it, but this can be overridden in their > user interface. Indeed it can, but one reason for inserting the chosen DOCTYPE is to inform anyone considering validating the page that they should not seek to override it in the user interface other than for reasons of pure intellectual curiousity. > Beyond that, it never had any impact, *except* that it triggers > "standards mode" in browsers - something quite different from its SGML > meaning. I would argue that it /does/ have in impact, but in the world of human beings rather than the world of computers and automata. It make a statement that the author of the web page claims to have some familiarity with the differences between various DOCTYPEs and has elected to use that specific one for reasons that he or she does not seek to explain (unless a human-readable comment follows which explains why that particular DOCTYPE was chosen). The fact that that statement may be false more often than it is true does not detract from its putative significance. > Most importantly, browsers do not care the least about "HTML versions" > (apart from text/html vs. XML media types, but this does not depend on > the DOCTYPE string). That is a very sad but true fact. > If someone alleges that your page contains errors, then a DOCTYPE is > just a formality. For example, the following document passes validation, > but it is neither conformant nor sensible: > > <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01//EN" > "http://www.w3.org/TR/html4/strict.dtd"> > <title/I made this up/ > <h6><img src=logo.gif alt="TURN IMAGES ON YOU FOOL"></h6> > > (The page uses an SGML construct that is valid in HTML 4.01 but not > supported by any browser. It also contains an alt attribute that is not > an adequate replacement for an image, violating HTML specifications, but > SGML validation cannot possibly detect this. And using a 6th level > heading as the only heading is just mad, but conforming.) As we both know, "[validation] can be used to show the presence of errors, but never to show their absence!" (with apologies to Prof. Edsger Dijkstra[1]). Philip Taylor -------- [1] "Program testing can be used to show the presence of bugs, but never to show their absence!" : Dijkstra, E W (1970), in "Notes On Structured Programming", corollary at the end of section 3, "On The Reliability of Mechanisms".
Received on Wednesday, 27 November 2013 19:41:25 UTC