Re: safe PUT

Larry Masinter:
>I also believe that there is sufficient negative
>reaction to Roy's proposal in the HTTP/1.1 draft to the '5 second
>delay' as to indicate that there is not 'rough consensus' on that
>particular design element.
>We've had a long discussion with various counter-proposals floated,
>but not much convergence. I'll ask all of you to decide between one of
>the following three alternatives (I can think of no others):
>- Are you interested in drafting a counter-proposal?


>- Should PUT go into HTTP/1.1 as originally specified,

Yes, the two-phase send scheme with 5 second timeout should be taken
out.  Not only for PUT, but also for other methods like POST.

> but with a warning as to its unreliability?

The warning should not be about unreliability, but about _limitations_
in the PUT semantics that make the implementation of certain types of
reliable services by using only one PUT impossible.

The proposed two-phase PUT does not offer a clear enough improvement
over the limitations of the current one-phase PUT scheme.  Without
such a clear improvement, we cannot to burden software authors with
the requirement to provide implementations of the new scheme.

>- Should PUT be removed from HTTP/1.1?

I don't care much either way.  If it is removed, then the reason
should not be that it is unreliable, but that is is not currently used
in any standard way.


Received on Saturday, 6 January 1996 09:09:58 UTC