[Top] [All Lists]

IPV6: Where are we? (SMTP)

2008-04-06 19:37:50

I hope I can get an honest discussion here about IPv6 and hears IETF and mail system overall thoughts on the subject. I know it is all subjective and will probably get different feelings here and among other groups, but unless I am off base, I think it might be healthy to discuss it here if only to serve as an increasing awareness and to help in the adoption process.

Frankly, I did more in the last 3-4 days learning about the status of IPv6 and from I am seeing, it seems to be the epitome or height of IPv6 was around 2005 and most of the work and implementation in outside the US.

I have seen a few presentations that suggest or summarizes the following highlights or key points:

  - the growth rate of IPv6 address assigned is higher in the US,

  - but the usage of it is very low in the US.

  - most of the work is in academia, startups or vendor R&D labs,

  - while many of the big brand name vendors have IPv6 web sites,
    they are not marketing it or effectively put it on the back
    burner.  Points were made that the clients and infrastructure
    is simply not ready and other points where made that there
    is a higher overhead with an IPv6/AAAA framework.

  - However, other points were made that ISP are also adopting
    it to help with addressing and bandwidth issues and they
    feel comfortable that this is acceptable with their market
    of customers.  A point was made that since IPv6 is backward
    compatible, this has worked very well for ISPs.  I think they
    meant that they have IPv6 to IPv4 tunneling software that is
    high quality and works very well for them.

  - Only a few SMTP servers support IPv6.  Some of the current
    name brands require a recompile to support IPv6 sockets.

  - I saw a point that doing a A first, then AAAA lookup for
    SMTP clients was discovered to be more OPTIMAL because
    there is a high failure rate of AAAA records.  Some of
    this failure was pointed at buggy DNS server which may
    report erroneous SERVFAIL errors when a AAAA query
    is done first, but not when a A record is done first.

  - I saw a related point which suggest that the socket level
    2 function getaddrinfo() be "hacked" or changed to avoid
    doing AAAA requests.  Overall, the recommendation was
    to avoid doing AAAA requests as much as you can.  I was
    not sure if this was for VISTA only or a general

  - Another related point is how doing a AAAA or A lookup
    will be masked by the getaddrinfo() request. So that
    might alter some logic in what is expected or not
    expected in SMTP.

Other items I have learned that I have not seen discussed here or missed:

  -  RFC 2874 or the A6 resource record.

     "DNS Extensions to Support IPv6 Address Aggregation and

Why has this not been discussed as an alternative to AAAA? It would seem to me that A6 which has type value of 38 is a reasonable item to be discussed if we also going to discussed lookups for AAAA records.

Overall, I think IPv6 is an overwhelming concept which have not yet put in enough time to consider for 2821bis, even if its just about one part regarding IMPLICIT MX. Are we going to make the right decision?

We know one thing - SMTP IPv4 operational behavior with a score of years of time tested engineering behind it.

I vote to NIX the whole IMPLICIT MX consideration for IPv6 for 2821bis and I hope the IETF cogs and a champion among us can take us into the next era with a new IETF chartered WG effort to see how to move SMTP into the IPv6 era, or even the feasibility of this effort. Maybe that was already done, I don't know. But it sure sounds to me, this is big mountain to climb for many.

Of course, this is one person's opinion only.


Hector Santos, CTO

<Prev in Thread] Current Thread [Next in Thread>