I think that the right solution, if we could redo things, would
be to allow something like  in content, but to never use
the actual byte values. This would allow the data guys to stream
stuff, and could leave the document guys reasonably unconcerned.
Of course then pattern restrictions on mixed content (which we
currently don't have) would become really helpful.
Regards, Martin.
At 08:07 01/07/18 -0700, Mark Davis wrote:
> > I wouldn't want any control codes in a database. Having a control-G
> > may be funny (the joke as I know it goes back to Don Knuth), but
> > something like a control-S is too much of a risk.
>
>*You* wouldn't want?
>
>There are a lot of characters *I* wish were not in databases, or in use at
>all. A lot of them may or may not make sense. Whether or not I want them,
>someone can have a database where they are allowed. By having this
>(inconsistent) restriction, it simply means I can't be guaranteed full
>round-tripping from databases to XML and back, no matter what their
>content.
>
>Of course, this is not a huge restriction -- it is simply a gratuitous
>annoyance. One could even live with something much more onerous, say XML
>disallowing all characters whose code points were divisible by 4321 -- just
>have complicated DTDs and shift into base64 if you encounter any of those
>codes.
This archive was generated by hypermail 2.1.2 : Thu Jul 19 2001 - 06:40:55 EDT