>
> Server's generally don't do this either, probably for speed.
I don't know about generally, but I believe this is a standard
behavior of the CERN server, at least until the buffer is filled.
> Currently, when a server break the connection, the client assumes this
> to be the end of the data stream. I don't know if they complain about
> any mismatch between a content-length and the actual length of the data.
Yes, and I'm experiencing a situation where that brain dead approach
has apparently left a partial docuemnt in a proxy that it can't
tell from the whole thing.
Content length should be required UNLESS another method such as
the packet approach under discussion is used to reliably define
the end of data. IN that case an estimate would be useful so that
friendly User Agents can offer an indication of progress, etc.
Dave Morris