Re: [Assimilation] Travis-CI and Assimilation

2015-12-03 Thread Alan Robertson 
We support ipv4 addresses but only by using ipv6. So, we can't
(currently) run on an ipv4-only platform. I couldn't imagine any such
platforms existing in the last 10 years or so.

I don't care if you can't route or send to/from "real" ipv6 addresses.
But internally, everything in Assimilation is an ipv6 address. When I
want to send to 10.1.2.3, I send instead to :::10.1.2.3. So, we
support ipv4 addresses - but we translate them to ipv6. On the wire, our
packets go out as ipv4 packets, not ipv6. But if you disable ipv6 in the
kernel, then we choke because that translation code is part of the ipv6
stack.

The vendor who doesn't support ipv6 is Google, actually ;-).  And
Travis-CI inherits that from them...

I find it really hard to imagine that Google built anything that was
ipv4-only. For 10 years it hasn't made sense. Maybe you should talk to
them ;-)

I could put code at the bottom end of my network stack that would figure
out that an IPv6 address was ipv4 and then send it as ipv4. But that
would be ugly. As far as I know, it would only be needed for Travis-CI
(Google).

Until today, I had no idea that the the "Google app cloud" was ipv4-only.

/me grumbles and thinks about putting the kludge into the code...


On 12/03/2015 03:08 PM, Jeff Silverman wrote:
> People,
>
> For the foreseeable future, both IPv4 and IPv6 will have to be
> supported by all vendors.  It's that simple.  You have to support IPv4
> because at the present time, there are devices that only support IPv4,
> and that will be the case for years, perhaps decades to come.  You
> have to support IPv6 because the planet is running out of IPv4
> addresses.  Everybody in the computer business in any capacity either
> knows this or should have known this.   I would amazed if anybody
> argued with my position on this issue.  That's not the same as
> actually doing something about it, but nobody - nobody! - disagrees
> that supporting both protocols is best practices and all new systems
> should do so.
>
> So if you have a vendor who is unwilling or unable to support both
> protocols, then I would drop them.  If Assimilation does not support
> both protocols, then that is a major, serious bug and should be
> addressed ASAP.
>
>
> Jeff
>
>
> On Thu, Dec 3, 2015 at 1:06 PM, Alan Robertson  > wrote:
>
> Hi Atom,
>
> I'm aware of that they are limited by their choice of suppliers,
> but they didn't ever mention ipv6 was going away until about 5
> days ago. I don't read their blog, so I didn't know until our
> builds started breaking randomly.
>
> If they supported ::1 (loopback), our testing would work just fine.
>
> They made a choice to eliminate this - and IMHO, a week is
> insufficient notice, and a note on a blog is an inadequate way to
> let people know.
>
> We now have to either wait for them to decide to keep a little old
> infrastructure around, or move somewhere else.
>
> From the perspective of the kernel, we don't support ipv4. It's
> something that would take a while to fix, and it's an incredibly
> stupid way to spend very limited resources.
>
> We can do everything we want to do using ipv6 - except test under
> Travis-CI :-(.
>
> -- Alan
>
>
> On 12/03/2015 01:52 PM, Atom Powers wrote:
>>
>> TravisCI can't really do anything until Google does something.
>> https://cloud.google.com/compute/docs/networks-and-firewalls#networks
>>
>>
>> On Thu, Dec 3, 2015 at 12:34 PM Alan Robertson 
>>  wrote:
>>
>> Hi,
>>
>> Travis-CI (continuous integration) has switched their build
>> infrastructure over to an /ipv4 only/ setup. They announced
>> this a few days ago:
>> https://blog.travis-ci.com/2015-11-27-moving-to-a-more-elastic-future
>>
>> The Assimilation software is ipv6-only - so this isn't a
>> match made in heaven :-(.
>>
>> I've created a Travis-CI GitHub issue for it here:
>> https://github.com/travis-ci/travis-ci/issues/5200
>>
>> I have no idea what they'll do about this - if anything...
>>
>> -- 
>>
>> Alan Robertson / CTO 
>>
>> al...@assimilationsystems.com
>> / +1 303.947.7999
>> 
>>
>> Assimilation Systems Limited
>> http://AssimilationSystems.com
>>
>> Twitter  Linkedin
>>  skype
>> 
>>
>> ___
>> Assimilation mailing list - Discovery-Driven Monitoring
>> Assimilation@lists.community.tummy.com
>> 
>> 
>> http://lists.community.tummy.com/cgi-bin/mailman/listinfo/assimilation
>> http://assimmon.org/
>>
>> -- 
>>

Re: [Assimilation] Travis-CI and Assimilation

2015-12-03 Thread Jeff Silverman 
Alan,

You're quite right.  I used to work for Google, and I had this argument
with my co-workers.  Their response was "Jeff, you are absolutely correct,
we really do need to implement IPv6, if nowhere else than at the border".
They told me to write a ticket to fix the problem, and the ticket was
rejected because there was another ticket on the same issue, so somebody
else already thought of it.

Grr.

So, the next step is to send an E-mail to Larry Page?


Jeff


On Thu, Dec 3, 2015 at 2:28 PM, Alan Robertson  wrote:

> We support ipv4 addresses but only by using ipv6. So, we can't (currently)
> run on an ipv4-only platform. I couldn't imagine any such platforms
> existing in the last 10 years or so.
>
> I don't care if you can't route or send to/from "real" ipv6 addresses. But
> internally, everything in Assimilation is an ipv6 address. When I want to
> send to 10.1.2.3, I send instead to :::10.1.2.3. So, we support ipv4
> addresses - but we translate them to ipv6. On the wire, our packets go out
> as ipv4 packets, not ipv6. But if you disable ipv6 in the kernel, then we
> choke because that translation code is part of the ipv6 stack.
>
> The vendor who doesn't support ipv6 is Google, actually ;-).  And
> Travis-CI inherits that from them...
>
> I find it really hard to imagine that Google built anything that was
> ipv4-only. For 10 years it hasn't made sense. Maybe you should talk to them
> ;-)
>
> I could put code at the bottom end of my network stack that would figure
> out that an IPv6 address was ipv4 and then send it as ipv4. But that would
> be ugly. As far as I know, it would only be needed for Travis-CI (Google).
>
> Until today, I had no idea that the the "Google app cloud" was ipv4-only.
>
> /me grumbles and thinks about putting the kludge into the code...
>
>
> On 12/03/2015 03:08 PM, Jeff Silverman wrote:
>
> People,
>
> For the foreseeable future, both IPv4 and IPv6 will have to be supported
> by all vendors.  It's that simple.  You have to support IPv4 because at the
> present time, there are devices that only support IPv4, and that will be
> the case for years, perhaps decades to come.  You have to support IPv6
> because the planet is running out of IPv4 addresses.  Everybody in the
> computer business in any capacity either knows this or should have known
> this.   I would amazed if anybody argued with my position on this issue.
> That's not the same as actually doing something about it, but nobody -
> nobody! - disagrees that supporting both protocols is best practices and
> all new systems should do so.
>
> So if you have a vendor who is unwilling or unable to support both
> protocols, then I would drop them.  If Assimilation does not support both
> protocols, then that is a major, serious bug and should be addressed ASAP.
>
>
> Jeff
>
>
> On Thu, Dec 3, 2015 at 1:06 PM, Alan Robertson  wrote:
>
>> Hi Atom,
>>
>> I'm aware of that they are limited by their choice of suppliers, but they
>> didn't ever mention ipv6 was going away until about 5 days ago. I don't
>> read their blog, so I didn't know until our builds started breaking
>> randomly.
>>
>> If they supported ::1 (loopback), our testing would work just fine.
>>
>> They made a choice to eliminate this - and IMHO, a week is insufficient
>> notice, and a note on a blog is an inadequate way to let people know.
>>
>> We now have to either wait for them to decide to keep a little old
>> infrastructure around, or move somewhere else.
>>
>> From the perspective of the kernel, we don't support ipv4. It's something
>> that would take a while to fix, and it's an incredibly stupid way to spend
>> very limited resources.
>>
>> We can do everything we want to do using ipv6 - except test under
>> Travis-CI :-(.
>>
>> -- Alan
>>
>>
>> On 12/03/2015 01:52 PM, Atom Powers wrote:
>>
>>
>> TravisCI can't really do anything until Google does something.
>> https://cloud.google.com/compute/docs/networks-and-firewalls#networks
>>
>>
>> On Thu, Dec 3, 2015 at 12:34 PM Alan Robertson 
>>  wrote:
>>
>>> Hi,
>>>
>>> Travis-CI (continuous integration) has switched their build
>>> infrastructure over to an *ipv4 only* setup. They announced this a few
>>> days ago:
>>> https://blog.travis-ci.com/2015-11-27-moving-to-a-more-elastic-future
>>>
>>> The Assimilation software is ipv6-only - so this isn't a match made in
>>> heaven :-(.
>>>
>>> I've created a Travis-CI GitHub issue for it here:
>>> https://github.com/travis-ci/travis-ci/issues/5200
>>>
>>> I have no idea what they'll do about this - if anything...
>>>
>>> --
>>>
>>> Alan Robertson / CTO
>>>
>>> al...@assimilationsystems.com / +1
>>> 303.947.7999
>>>
>>> Assimilation Systems Limited
>>> http://AssimilationSystems.com
>>>
>>> [image: Twitter]  [image: Linkedin]
>>>  [image: skype]
>>>