> Why resolve a Host name when you know the actual IP address? It will
> Eliminate any translation , thus finding its way to the stack faster.
> It does however have to travel to the NDIS layer, and back which would
> account for some speed loss, but certainly not 10 seconds.

The Microsoft TCP/IP stack was added later and I don't think it is right even in
W2k. We are not talking IB here. The machine name is always looked up in HOSTS
first - even if it is a raw address - it is that lookup that then returns the
either the real raw address - or the machine name. SO if you feed in a raw
address normally you get one back, which is then identified as an address and
used. If you don't, then DNS and other lookup's take place, using the machine
name. I think that if it does not find a match in HOSTS then there is a delay
before it gets the return - it depends on just which version of TCP/IP is
installed.

> > The golden rule is - although you CAN, don't use a raw IP address to
> connect. I don't know why this recommendation exists but it's documented
> and the archives show a lot of people reporting problems from it.
> >

See above - In Win98 the raw address may work or may not - NT and W2k seems to
be OK but there is a delay when there is no match in HOSTS.

You can try adding
192.168.0.2 192.168.0.2
to HOSTS

This speeds things up on some setups, but on others it will not work at all. I
had thought the make of network card came into it, but even where we have 50
identical machines, a couple have problems in this area - the machine name will
not work and we have to use raw addresses.

--
Lester Caine
-----------------------------
L.S.Caine Electronic Services