Re: note from Prof Knuth

Dear Olivier,

Thank you for your well-thought-out reply to my flame.
I know that creating such a reply must have taken a nontrivial
amount of time from your schedule, and I know also that you are
voluntarily contributing your time to W3C. So I don't want to add
unnecessarily to your burden.

I am sure that 99% or more of all web pages make use
of conventions that came out after 1995. But I still believe
that the absolute number of legacy pages is huge, although
not high percentage-wise, even if we count them by current hits
rather than by their existence. (Similarly, for example, I
would not tell numerical analysts to forget about using computers
any more, because so few computers nowadays are devoted to numerical
analysis. Their percentage keeps shrinking, yet the absolute
numbers and the quality keeps growing.)

I didn't know that Netscape's DTD wasn't ever "standardized" in
any way, not even published. But I am sure that it is "widely
supported"; certainly I've seen my pages handled properly by
dozens of browsers as I've traveled around the world, and
never handled improperly. At least not since 1996 or so.

After receiving your note, I looked for about half an hour at
"Cascading Style Sheets". The whole idea of such things
has always been scary to me, because I haven't time to
learn every new language that comes along. I find it far
more efficient to stick to a few basic tools that let me
get quality stuff out the door quite efficiently. Certainly
if I ever do use such a thing, it won't be by applying somebody's
"tidy" system software to my documents; then I'd have to upgrade their
tool every time a new feature came out, etc etc. Instead I would much
prefer to do everything myself in a way that I understand and that
I believe will last another twenty years (by which time I'm
probably dead). When I read notes like "put your style info
into HTML comments, because some browsers don't understand them",
you might imagine why I'm a bit chary --- since I know that
the conventions that I have been using, for more than 3000 days,
DO work! The reason I've kept using oldfashioned stuff on my
web pages was precisely because I knew it was working.
Only the validation service tells me otherwise, and only since
a short time ago.

I've got about 5000 occurrences of "absmiddle" on various
web pages. Those pages fall into dozens of different categories,
and a zillion special cases arise; thus, I see no advantage
of having a style sheet approach in my case, at least not
with respect to a whole bunch of different pages. I could
maybe add style information at the top of every page,
in order to economize further down.

I think I understand the reasoning in your message except
for one thing: Why could it possibly be right to insist on
the "alt" attribute always being present, instead of letting
alt="*" be the default?? You said you wanted to save bandwidth.
But I'd have to add many thousands of characters to my pages,
just to put in a default that every browser obviously already
has built in.

I realize that accessibility is a great virtue, etc. But do
you guys need to legislate morality? When I put up an image,
I do ask myself "is there a useful alt?", but almost always the answer
is "no". And indeed, all the normal reasons for having a default
are present. Certainly in a "loose" standard DTD, at any rate.

You suggested that I try
in place of
But I just tried it, and got the error message
  there is no attribute "valign"
It was curious, though. My source file said
  <img src="gaoduhnah.gif" valign="middle"
   border=0 alt="U+9AD8+5FB7+7EB3">
and the error message put the quotemark before middle in red,
as if indeed there was no such attribute; yet later on the same,
page I said
  <img src="/cgi-bin/count.cgi?df=knuth-index.dat|sh=0"
      alt="*" valign="middle">
and there was no error message! (If you too are curious, try
and seeing what happens.)

If you can tell me something that works for align="absmiddle"
with your transitional 4.01 DTD, I will probably switch all
of my hundreds of pages to that DTD. You are right, it
won't take a week. I was flaming.

Also, I suspect you could tell me how to insert a few lines
of CSS code so that your validator automatically appends
alt="*" to every img that doesn't already have an alt attribute.
(I don't understand why I should actually
put this code on every page, because it will just make every
browser run a bit slower without changing the formatting
behavior in any other way. But I may do it, just so that I
don't have to wade through so many error messages when I
try to validate pages in future.)

Thanks for listening! -- Don Knuth

Received on Tuesday, 20 September 2005 23:24:08 UTC