> I'm not talking about sending anything and wasting bandwidth. There
>are extensions that are useful to all users and extensions that are
>only useful for specific situations. The server knows that better than
>the client. I would consider a client that breaks down on unsollicited
>extra data not well implemented. As a user I would prefer a client that
>allows me to see the unsollicted extra data. But my client wouldn't say
>to the server that it accepts anything because the client can not judge
>the usefulness for the user. The server can.
I don't see how the server, written to serve data to any client, can
know who the user of each individual client that connects to it is,
better than the actual client the user is using.
The model for SRU is that because the interaction is (relatively)
straight forward between client and server, it is much easier to write
clients for specific purposes, and hence there should (in theory) be
many more client implementations than server implementations. With
Explain to describe what features the server makes available, the client
can then configure itself and makes requests as needed.
Are you debating that this is the case?
If you're not, how is it that the few servers can better understand the
users (or perhaps NO human user for metasearch applications) than the
many clients written for specific purposes, which have access to the
server's explain record?