once again about GDB protocol

[Andrew Cagney]

> Hi,
> 
> It seems like any improvements in this area are impeded mostly by
> 'ooops we have a bad implmentation in the field which will broken by
> this nice feature'.
> 
> Does it seem reasonable in this case to introduce a different framing
> character (e.g. '@' instead of '$') for the next/improved version of
> the protocol ?

Given that one of the bugs is the entire transport layer (packet and 
ack) a new packet is needed.

As for the actual framing character, I've no opinion.

> So, we can have a new/shiny/fast/smart/functional protocol and when we are
> connecting to an old stub we will simply use 'old-remote' target, it seems
> a trivial feat to make both 'remote' and 'old-remote' target to bark a
> meaningful meesage about protocol mismatch and about necessary actions to
> be taken to correct it.

yes.
Andrew