Re: END_SEGMENT and headers, #2

On 2014–04–19, at 4:03 AM, Martin Thomson <martin.thomson@gmail.com> wrote:

> On 18 April 2014 12:02, David Krauss <potswa@gmail.com> wrote:
>> I’m designing an interface and can’t bring myself to expose which type of frame defined the segment boundary. END_SEGMENT in HEADERS will just look like an empty data frame.
> 
> I'm not sure that I would expose segments myself.  Depends on whether
> you have an actual use for them.  

It’s for a generic library which is intended to support any protocol that could be layered on HTTP/2.

> Breaking a read() or event when you
> see END_SEGMENT or any HEADERS frame would probably work if your
> intent is to maintain segment breaks.

Yes, it seems that HEADERS practically requires END_SEGMENT semantics whether or not the bit is set. The bit is then only informative to the application which can query it, but what are the semantics of that corner-case query?

Not to beat a dead horse, but I think that bit should go. Data frames alone provide exactly enough END_SEGMENT flags.

Received on Friday, 18 April 2014 20:13:10 UTC