Re: Design Decisions and Hurd/L4 work (was: Re: Improving Hurd)

[Thomas Bushnell, BSG]

Jan Atle Ramsli <trollet@skynet.be> writes:

> The problem is that someone just popped out and said "The ADT is not
> longer relevant, it died 10 years ago".

That's not what I said.  You are getting really annoying; it's as if
you *want* to misunderstand.

> That is: I have been told that some problems can not be solved with the
> concept of an ADT.
> I need to know: What are they. It takes only one example to convince me.

Geez, an elementary programming text might help.

Two classic examples:

First failure mode: ADT specifications do not include information on
time complexity or expected uses.  Consider for example an X 11
"window" as an ADT.  Now, what would you say if a spreadsheet design
started off with "ok, create 1,000,000 windows, one for each cell, and
we can just use X to manage display and positioning".  Whoops!

Second failure mode: It's not possible in a strict ADT style to have
objects which live on several efficiently managed data structures at
once.

Thomas