[ih] Date of RFC 791 for celebration

[David P. Reed]

Yup, I remember the battle about variable length addresses well.  One 
that I was on the losing side of, since the folks doing implementations 
said it would be enormously expensive to have fields that didn't end on 
32-bit word boundaries.

Noel Chiappa wrote:
>     > From: Jack Haverty <jack at 3kitty.org>
>
>     >> On Wed, 2006-03-29 at 09:13 -0500, Noel Chiappa wrote:
>     >> So it looks as if my prior take:
>
>     >>> 3 was the first version that had the headers fully split ... but my
>     >>> guess is that it included variable-length addresses. I seem to recall
>     >>> that 3.1 had the variable-length addresses removed, and 4 was an
>     >>> editorial cleanup of 3.1
>
>    .>> is likely accurate (although I wish we could recover a 3.1 spec to be
>     >> sure).
>
>     > I don't think it's right to think of 4 as an "editorial cleanup" of 3. 
>
> Ah, I didn't say it was a cleanup of 3, I said it was a cleanup of 3.1! The
> two were (I'm fairly sure) very different - TCP/IP-3 had variable-length
> addresses, and I'm pretty sure that 3.1 had the fixed-length 4-byte addresses.
>
>
> TCP/IP 3 really did have variable-length addresses. Here's an excerpt from
> IEN-21, "TCP 3 Specification" (which also contains the IP specification); it
> says, in section 4.3.1 "Internetwork Packet Format":
>
>   An address is a variable length quantity (in multiples of octets). It is
>   intended for the first octet of an address to be interpreted as a network
>   identifier, and that the rest of the address identifies a host within that
>   network.
>
>   ...
>
>
>     DAL: 4 bits
> 	Destination Address Length in octets.
>     SAL: 4 bits
> 	Source Address Length in octets.
>     Destination: variable
> 	The destination address, DAL octets in length.
>     Source: variable
> 	The source address, SAL octets in length.
>
> Figure 4.3-1 in that document shows this all too.
>
> 	Noel
>
>
>