Re: associating constant feature set

At 09:30 20/01/99 -0800, Bill Newman wrote:
>[...] I myself
>would prefer to separate the idea of indirect references (which is in
>my mind the same as what I've been calling delegation) from the idea
>of dealing with defaults.

Good principle, I think.

>Indirect references (without defaults) could be useful for many mobile
>devices. Defaults (without indirect references) could be useful to
>solve the font problem. And the two features work well together, too:
>defaults increase the power of indirect references.

... and that's a good example why

>I think both problems are important -- I actually worry more about
>defaults and the font problem than I do about indirect references.
>But I also think indirect references are easier to solve. Extending
>the algebra to allow just the right beyond-first-order-logic things to
>be done, and doing it in a flexible, easy-to-understand,
>not-too-hard-to-implement way is subtler and probably harder.

I suspect these are both difficult problems, with different kinds of subtlety.

IMO, the "default" problem is relatively self-contained so a solution can
be developed and tested in relative isolation.

OTOH, I se the "indirect reference" problem as something that (potentially)
impacts/interacts with the behaviour of other protocols (including transfer
protocols), and that begs questions like:  how to maximize othogonality
between representation and mechanism?;  how to ensure the solution is
scalable for tjhe Internet?;  etc.

To recap, I think the separation of defaults and indirect references is an
important insight -- thanks for that!

#g

------------
Graham Klyne
([email protected])