Re: Multiple Alt-Svc parameters of the same type

[Martin Thomson <[e-mail]>]

On 19 January 2017 at 23:00, Lucas Pardue <[e-mail]> wrote:
> Is it correct to interpret this statement to say that Mike's definition of the v parameter (or whatever it turns to be called) is a specialisation of the general case and it's "ordered list" processing is implemented at the required level.


Mike's definition of v= in the quic documents defines the semantics
for a field in a specific context.  That definition, in addition to
specifying semantics, can also require that the list form be
understood by implementations.

Another definition for a different tuple of header field and parameter
could also insist in a different definition: maybe that multiple
instances of the same name result in an error of some sort.

It would be good if the common syntax established what is possible.
I'm thinking:
1. multiple values are possible

2. a definition for a specific parameter may forbid the use of
multiple values, (noting further that a sender might not be able to
guarantee that the receiver knows of these rules without some sort of
prior indication that the semantics are known)

3. order {is significant|is not significant|has significance as
defined by the field semantics}

4. where multiple values are accepted specific rules should be
described for handling errors caused by a single parameter, this could
render many different things invalid: just that parameter, all
parameters with the same name, the affected header field value (there
could be multiple comma-separated values), or the entire header field.
By default, only the parameter value is affected.

Or something like that.  I might have missed something.