Re: Re: smart upload server
On Mon, May 24, 2010 at 6:03 PM, Petr Jašek <firstname.lastname@example.org> wrote:
> I'm thinking of PUT for files and GET for commands. Or maybe combine both
> two and use just POST, which could make easier implementation for other
> transport protocols.
> What do you think?
I am happy with PUT and GET. I hope there are no broken proxies in
real life that block PUT.
>> Is the protocol stateless? How does the client reference the
>> changes file in the second and later requests?
> Stateless if possible, with some time window for each changes file,
> checksums are on server so he knows what should come.
I see. But it might be more difficult with partial uploads. I suggests
to include the checksum of the changes file as a header (e.g.
X-Changes-Sha256: ...) - even for the upload of the changes file
itself. We have seen aborted uploads of changes files in the past.
Having the checksum in the header would make it easier to detect such
problems at least for less advanced implementations of the server.
Feel free to ignore the checksum header if you don't need it in dak's