Re: [rohc] TCP/IP EPIC profile

["West, Mark (ITN)" <[e-mail]>]

Hi Julije,

I've just waded through your comments on the 'expanded' profile, and 
I'll try and summarise the points you raise and make some comments...

- IP checksum treated later / INFERRED-IP-CHECKSUM not defined

The current draft does not define INFERRED-IP-CHECKSUM.  (It was defined 
in an older 'EPIC for TCP' draft, but I realise that's not much help!) 
It is in the new version that will be submitted this week, though.
When you see how this works, you will appreciate that the IP checksum is 
not actually treated later, but simply ignored.  (On compression, 
INFERRED-IP-CHECKSUM sets the checksum bits to 0; on decompression is 
fills them in with the computed checksum).


- NBO not in the toolbox

Again, true.  We have removed INFERRED-SCALED and replaced it with NBO 
and SCALE operations, which gives more flexibility.  Again, these are 
described in the draft that will be submitted.  (I agree that it might 
have been nice to synchronise things a little better...)


- Some percentages don't add up to 100%

That's what we profile writers refer to as a 'mistake' ;-)
Actually, it shouldn't matter - but it should be fixed!


- Compiler should read comments or what?

Well, the compiler is free to read the comments.  But it probably ought 
to ignore what they say...


- Various mis-named method names

Sorry - these are typos and need to be fixed.


- Various unused methods

Again, they should be referenced so this is an error in the profile.


- FORMAT appears in 'CO' packet

Yes, it certainly does.  Otherwise, how would you know what different 
formats to build?  I don't quite see the problem here...


- ack STACK-TO-CONTROL not defined
- psh STACK-TO-CONTROL not defined

I don't understand the comments.


- skip-msn is not defined (but self-explanatory)

Tend to agree!  (This is, of course, a TCP specific method and should be 
defined along with the profile).


- An IRREGULAR(16,100%) is used to encode a format selector

This should be IRREGULAR(1,100%) - it's a typo...


- Not clear what is being encoded at each point

This could be overcome by better commenting (whether or not the compiler 
reads them) or splitting up the methods to a finer granularity.  This 
should be taken care of in the profile, though, I agree.

Cheers,

Mark.


-- 
Mark A. West, Consultant Engineer
Roke Manor Research Ltd., Romsey, Hants.  SO51 0ZN
Phone [phone]   Fax  [phone]

(Yes, I do know that my disclaimer is in an attachment.  And, no, I 
didn't ask for it to be that way)