[Interest] TCP ACK with QDataStream -- OR: disconnect race condition detection

[Konrad Rosenbaum]

On Tuesday 11 September 2012 01:48:08 Constantin Makshin wrote:
> TCP's 16-bit checksum isn't [very] reliable by today's standards, so doing
> additional integrity checks at application level is not necessarily a bad
> idea (in situations where correctness of the transmitted data is REALLY
> important/critical).
> 
> On Sep 10, 2012 11:43 PM, "Konrad Rosenbaum" <konrad at silmor.de> wrote:
> > * You do not need to checksum them against physical defects.

A 16bit CRC is more than adequate for simple physical defects in todays 
networks. The probability that 2 bits flip in a way that keeps the checksum 
valid is extremely slim (smaller than a 16bit checksum would suggest at first 
glance).

I would be worried if Jumbograms of 1MB or larger were common. But we have an 
Internet with the average package being somewhat below 1500 bytes long and an 
ISP side MTU of 4kB.

If you need security against malicious attackers: use a cryptographic hash (at 
least SHA-1) and proper encryption (e.g. by using SSL).

If you transfer data in which every single bit is life-threateningly important 
and you use it only long after transfer: use SHA-1 checksums and store them 
with the file. Thre probability of corruption while in storage is at least as 
high as the probability of corruption during a TCP transfer. I would also 
recommend highly redundant hardware with at least two independent 
implementations of the software in this case. ;-)



	Konrad
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.qt-project.org/pipermail/interest/attachments/20120911/c8c08e69/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.qt-project.org/pipermail/interest/attachments/20120911/c8c08e69/attachment.sig>