On 03/15/2012 04:00 AM, Pedro Alves wrote:
>> > This patch series doesn't include the patch of IPA protocol part. I am
>> > sure there should be some changes in protocol during the review, and
>> > I'll post it when these patches are reviewed.
> As I said before, I'm not convinced the extra trouble of having yet
> another protocol is really a good idea, so I'll just not comment on it much,
> except to note that without a protocol version in the stream somewhere,
The only one use case about protocol version is that some command may
vary on different versions. When see a command in stream, it should be
interpreted differently according to its version. Anything else that we
need a version of protocol?
> it seems you're likely to have trouble latter on when we want to add
> extra optional features.
Hmmm, we have a 32-bit flag, and each bit in this flag represents the
agent whether or not has a capability or feature. AFAICS, features are
boolean state, so one bit for each should be fine. Am I missing