100% - 80% = ?

Thomas Dowling tdowling at ohiolink.ohiolink.edu
Thu Mar 7 09:37:15 EST 1996


My organization is now testing or using production (!)
versions of web interfaces from 5 different systems or
database vendors, and not one of them reliably puts out
HTML which could pass even a comparably forgiving 
validator like WebLint.  And I'm not talking about little
things like leaving out <link rev="made"...>, or even more 
major stuff like having a <HEAD> and a <BODY>.  I consistently
see botched tagging which I know leads to unpredicatable or
completely broken results with some browsers.  Examples below.

The annoying thing about this is that I also know from 
talking with several of these vendors that they are only
"checking" their work by looking at it with Netscape, and
probably only Netscape 2.0.  We're like a lot of sites in
that we see about 80% Netscape hits on our server (although
about half of that is from versions before 2.0).  But in
what other context would I be expected to go to one out of
every five users and say "Sorry, the system we bought 
won't work for you.  YOU have to change the tools you're
working with"?

I know we're not the only organization testing or paying
for these web products.  Is anyone else out there telling
vendors that their HTML has to validate?


Thomas Dowling
tdowling at ohiolink.edu


HTML HALL OF SHAME

Vendor's main database selection menu is presented as a 
<TABLE> which unfortunately has no <TR>, <TH>, or <TD>
elements.  Cheerfully rendered as a blank screen on two
of my browsers and a third can't figure out the anchors
within the table.  Aside from Netscape, only Lynx handles
this, because it ignores the table stuff entirely.

Vendor includes a background GIF, the leftmost 80 pixels of
which are black.  Attempts to move text off the black area
by prefacing the entire document with <DL><DT><DD><DL><DT><DD>.
HTML 101: just because a <DL> list is indented on your screen,
don't assume it's indented the same amount, or at all, on mine.
To duplicate the effect, put your thumb over the left inch of
this screen and try to read the rest of this message.

Vendor includes an inline image I provided, using it (as I do)
in a <p align="right">.  Vendor never closes the paragraph or
starts a new one, so an entire page is right-aligned.

<META HTTP-EQUIV="REFRESH"...> outside of the <HEAD>.  Brand-
new, so I don't know what all our browser will make of it, but
since the URL listed as the content for this is written wrong,
even browsers that try to follow it will get an error message.

Vendor's CGI script apparently does not preface output with a 
MIME declaration ("Content-type: text/html").  Reliably rendered 
as plaintext on at least one browser I commonly use.

Unclosed <img tag.  Unquoted alt text attributes.  <A HREF=...>
with no corresponding </A>.  Your guess is as good as mine about
what a browser will make of this.

Empty <a name=..."> tags.  A </ul> with no corresponding <ul>.
(better than the other way around, I suppose).  A </FORM> with 
no corresponding <FORM>.  Just indicators that they aren't
even doing rudimentary checking.


I could go on and on (and I have, to little avail).  



More information about the Web4lib mailing list