Re: A's content model

Terry Allen (
Thu, 22 Sep 94 11:30:04 EDT

| Let me try again. Suppose we have:
| <ul>
| <li><a href="#l1>src one</a>
| <li><a href="#l2>dest two</a>
| <li><a href="#l3>dest three</a>
| </ul>
| <ul>
| <li><a name="l1">dest one: legal and works</a>
| <li><a name="l2">dest two: illegal, but works
| <li><a name="l3"></a>dest three: legal, but doesnt work
| </ul>
| You will find:
| link works with Mosaic parses by current DTD
| 2 YES NO
| 3 NO YES

Yes, this is what I found, and why I brought the issue up (though
I hadn't caught that 1 will work; I tested 3 instead; this is
seriously strange).

But it is not acceptable that 2 will not parse, although it works.
While it's not too neat that 3 will parse but not work, that is
less important.

| I see no reason why link #3 shouldn't work. That it does not, I consider
| a bug in Mosiac.

Apparently more slipshod work by the Mosiac designers. Further
experiment shows that inserting a blank space or a line break within an
otherwise empty A makes no difference (although to SGML that counts
as content).

| I also see no technical reason why link #2 shouldn't parse. But
| it currently doesn't, and the sentiment is that it shouldn't,
| on the grounds that allowing folks to omit </A> is confusing.

2 doesn't parse because it doesn't have an end tag, which is
currently required (we both know this, just repeating it).

Notice that if the A has only a NAME att, it really doesn't matter
to the application (as distinct from the parser) what the content is;
that A is only marking a location and creates no hot spot.

We are describing current practice, which is confusing. I think we
ought to do it thoroughly, which would mean allowing end-tag
omission on A and warning users that case 2 won't work, even
though it parses. However, as this is indeed a bad bug in Mosaic
and time is short, I won't urge the issue further if no one else cares.

Terry Allen  (   Editor, Digital Media Group
O'Reilly & Associates, Inc.    Sebastopol, Calif., 95472