Hi Jonathan, 

I am ok with 2 or 3. 
Regarding compression of both ports simultaneously, I do not know if it
useful, but as we must support at least 3 scenarios (src only
compressed, dst only, or none), we need two bits. Hence I would say we
can reserve the 4th value for the case where both ports are compressed.

Regards,
Julien

-----Original Message-----
From: Jonathan Hui [mailto:[EMAIL PROTECTED] 
Sent: mercredi 8 octobre 2008 16:45
To: Julien Abeille (jabeille)
Cc: [email protected]
Subject: Re: Hc-01 LOWPAN_UDP encoding


Hi Julien,

Fixing UDP compression was definitely on the list of TODOs.

So there are a few options:

1) Expand the compressed range to 256 ports (the encoding lengths for
the ports are then 2, 3, or 4 bytes).

2) When compressing only one port, add a 4-bit pad as you suggest.  
However, I would rather add the 4-bit pad to be with the 4-bit
compressed port and keep the 16-bit uncompressed port byte-aligned (the
encoding options would be 1, 3, or 4 bytes).

3) A hybrid of the two: When only one port is compressed, the compressed
form is 8 bits (don't waste the 4 bits on padding). But when both ports
are compressed, both only 4 bits.
00 - Neither source nor dest are compressed
01 - Source port uncompressed, dest port 8 bits.
10 - Source port 8 bits, dest port uncompressed
11 - Source port 4 bits, dest port 4 bits.

Basic question to ask the group is if compressing both source and dest
ports to 4 bits simultaneously useful?

--
Jonathan Hui



On Oct 8, 2008, at 2:22 AM, Julien Abeille (jabeille) wrote:

> Hi Jonathan, all,
>
> one comment regarding hc-01 LOWPAN_UDP encoding for now source and 
> destination ports compression are splitted. It means if only one of 
> them is compressed, the UDP compressed fields will be 36 bits long 
> (4bits port + 16 bits port + checksum). We lose byte alignment this 
> way.
>
> to avoid this, you could either not split the compression of ports, or

> when only one is compressed, pad to 3bytes (4 bits port + 16 bits port

> + 4bits padding) What do you think?
>
> Cheers,
> Julien
>
>

_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan

Reply via email to