Single-byte I/O for keytab reads

[Ken Hornstein]

>I  would think that using unbuffered IO is fine provided that you read
>in large chunks.

I agree, but the whole structure of the keytab code is set up to read
in very small chunks.  So if it needs to be rewritten, then no problem ...
that's just a lot of work compared to removing a few setbuf() calls :-)

(Although I hadn't appreciated the idea of wanting to make sure that the
private key doesn't appear in a core dump).

--Ken