If you have/site has _that_ bad connection, use ssl. It provides yet another integrity check (Which really isn't neccessary, unless either site or client is having serious hardware/software issues... and in that case, there's no way you could trust the cheksum site would pass)
Quote:
Google:
> My initial question was about detecting random transmission errors in
> the payload (using checksums or by encoding the data...). I assume that
> Erlang just relies on the TCP and IP layer checksums?
If you can't trust your TCP stack, you have bigger problems than
that - your network-mounted filesystems (over SMB (TCP), NFS (TCP) or even
worse, NFS over UDP, go figure!) may be corrupted, so maybe you are not
even compiling the right source. Worse, your HTTP (TCP) or FTP (TCP)
download of Erlang and OTP might be corrupted! Even, your SMTP (TCP) mail
feed might be garbled! Maybe neither of us wrote this?!?
Joke apart: the odds of both the TCP CRC16 and the Ethernet CRC32
(or whatever link-level layer) ending up correct with wrong data are
*really* small.
Regards,
Miguel
|