Mike, what do you mean by "Real Stuff"? And what pollution has been
In general I think that making responses self-explaining - regardless
of the intelligence of the client - results in much more robust
applications  than when clients have to keep track of the session


>>> [log in to unmask] 25-3-04 15:22:11 >>>
> Date: Thu, 25 Mar 2004 12:29:33 +0000
> From: Robert Sanderson <[log in to unmask]>
> > Sorry to blend into an ongoing discussion, buuut I do not think
> > that a standard should aim at beeing good for brain-dead
> > clients.
> Thankfully, now we have the extension mechanism to test things like
> this out in practice before they make it into the protocol proper.

Yeeesss.  Extensions are certainly a less painful path to take than
constantly messing with the standard.  But I am still worried by the
encroaching influence of Dumb Clients on SRW.

Here's how the history of Dumb SRW/U Clients appears to me:

1. SRW/U invented.
2. People observe that you can make a neat demo just with a web
3. SRW/U extended to allow slightly neater demos.
4. (Reapeat step 3 several times)
5. SRW/U standard has got big and ugly, the packets sent over the
   network are twice the size a Non-Dumb Client needs, and Dumb
   Clients _still_ don't completely work because (stand by for a
   surprise) they're Dumb.

I fear that we are in danger of progressively polluting the protocol
in the pursuit of a series of increasingly marginal gains.  Sooner or
later people who want to do Real Stuff with SRW/U will have to bite
the bullet and start to write Non-Dumb Clients anyway, so I'm not
happy about continuing to make concessions to the dead-end street of
ever-fancier Dumb Demos.

And that's all I have to say about that.

/o ) \/  Mike Taylor  <[log in to unmask]>
)_v__/\  "That which gets lost in translation" -- Robert Frost's
         definition of poetry.

Listen to my wife's new CD of kids' music, _Child's Play_, at