[ntp:hackers] [Gpsd-dev] Single-writer/many-reader consistency

Eric Raymond esr at thyrsus.com
Sat Mar 26 22:14:40 UTC 2011


Dave Hart <davehart_gmail_exchange_tee at davehart.net>:
> If you do checksum, ISTR CRC32 isn't terribly strong for 8k.  MD5 is
> widely-available and reasonably fast, and for this purpose, I beieve
> its cryptographic weakness is any reason to avoid it.

On the assumption that you mean cryptographic strength is not a
requirement, I agree. 

I have a verified implementation of the 24-bit checksum that Qualcomm
designed for situations rather like this in my pocket, because one of
the DGPS correction protocols uses it.  It's faster than MD5 and
probably the right thing here.  PGP uses it.

Your message size is enough smaller for my Qualcomm implementation to
be overkill.  I agree that CRC32 is probably a good match for NTPD's
requirements.
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>


More information about the hackers mailing list