RE: Implementing NodeList

> I'll stand by my original statement.  For example, consider a document
> derived in the obvious fashion from a table in a relational database.  A
> node with several million children is not unreasonable to expect.  If the
> document is ``live'', the implementor might have to query the database and
> obtain the entire table at each iteration.
>
> It might not be obvious now, but people are going to be using the
> DOM for a lot of things besides hacking flashy effects into a document in
Netscape.
>
> > Nevertheless, the real answer -- as I think you said last time
> -- is to put interators back in the spec for Level 2.
>
> You're right there, but even iterators don't help if they are
> required to be ``live'' and constantly reflecting the current structure of
the tree.
> Sometimes that interpretation is the correct one, sometimes it isn't, and
> either the determination should be left to the implementor, or
> two different
> subclasses of iterators should be specified.

Inso often deals with document exceeding 10MB, and in fact, it is very
common for us
to deal with 100MB+ documents.

I can tell you that I was one of the people screaming very loadly about
"liveness".

That said, I think the easiest way to take care of liveness is to add a
locking
mechanism in level two. That way, you can *preclude* people from messing
with
the tree when you don't want them to: just build the tree, lock it, and then
liveness is a non-issue.

Received on Tuesday, 28 July 1998 11:38:40 UTC