- From: Balint Nagy Endre <bne@bne.ind.eunet.hu>
- Date: Mon, 13 Nov 1995 01:38:03 +0100 (MET)
- To: Shel Kaphan <sjk@amazon.com>
- Cc: http WG <http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com>
Shel Kaphan writes: > > I'm not going to say if I think byte ranges as presented here are A > Good Thing or A Bad Thing (since there seem to be some valid arguments > on both sides), but I would like to suggest that it might be > considerably safer if there was a standard way the client could > include in the request the last-modified date, or a checksum of the > whole document, that it last received in connection with the original > fragment of the document that it received. Then servers too could > help to keep things in sync, in particular by returning an error > status if the request and some available version of the document don't > match up. Except for a request for an initial fragment of a > not-yet-seen document, clients should probably always include this > information. To make this working, allways when a 'Partial content' response sent, we should require sending checksums and digests for the whole url and the part transmitted if the server supports any checksums/digests. > Regarding certain other parts of this discussion I have not really > read too thoroughly, the byte range should be defined in > terms of the *server's* original byte numbering, before any > end-of-line processing, double-byte character conversion, or whatever. > And intermediates shouldn't do any such processing. Or else this is doomed. The last 1.1 draft I read says something like: intermediaries and even clients should cache documents without any change in content, the only case when a client may and MUST apply conversions to local line-endings and other such stuff when it does a 'save as' operation. If my memory is correct, then we have absolutely no problems with byte-ranges. Andrew. (Endre Balint Nagy) <bne@bne.ind.eunet.hu>
Received on Sunday, 12 November 1995 16:50:14 UTC