[OpenAFS-devel] .35 sec rx delay bug?

[Buhrmaster, Gary]

> Isn't the TCP checksumming enough?=20

Interesting question.  And the answer is probably not.  I
can not find the reference right now, but with bulk high=20
speed transfers becoming common, the odds are against
the entire transfer not having at least one undetected
packet error is not on your side.  All the high energy
physics bulk data transfers now include additional
check summing simply because TCP checksums are=20
not sufficient (corrupted transfers).

> Anyhow, encryption would also have=20
> this effect.

Well, it depends on the encryption style, but for most
interesting encryptions, you are correct that the packet
would fail to decrypt if the packet had errors.