re: The Chinese Are Coming, Says PollHi Abby, Been there, done that, and the T-shirt still smells of garlic :-)
Actually I worked in the vendor community for the past 20 years or so, and only became an analyst a couple of years ago.
It's much more fun for an analyst to look at the results of a standards group's works, than the way it achieves those results. That way, instead of getting down into the bits, bytes and territorial disputes, I can make sweeping observations like...
"If we look at the goals of the MPLS Working Group and compare them to the original goals, we can see that the IETF has now hacked together MPLS into a working set of protocols that do actually allow connectivity to be achieved, but that additional MPLS protocols have yet to be developed to make this technology carrier grade."
BTW I base this conclusion on the two current goals of the WG, which are OAM and inter-area operation. These are sine qua non for carriers.
I've had more experience of the Frame Relay Forum and ATM Forum approaches than the IETF. IETFs are great fun, but I'm not sure my humming is up to the massed choirs of Cisco and Juniper :-) However, I did follow the famous OAM thread a couple of years ago. If that's an example of how a WG works these days I wouldn't want to advertise it if I were you :-)
Only kidding. I think it would be far more beneficial for a carrier engineer to follow your advice, since they are the ones who have to try and get this disparate bag of protocols to work at the end of the day.
In contrast vendor engineers are focussed on achieving a competitive advantage for their existing technology. Have you seen one of Yakov Rekhter's latest conference presentations on why some vendors prefer to evolve protocols, while others prefer to develop new protocols?
re: The Chinese Are Coming, Says PollHi stephenpcooke, Sorry to mis-attribute you :-) It had been a long day before I posted.
I totally agree with your perspective, that current IP standards are not sufficiently customer driven. And Peter Heywood presented an explanation earlier in the thread as to why that might be. To paraphrase Peter...Prior to deregulation, carriers used to be able to talk freely together at meetings because they knew they couldn't compete in each others' geographies. Today they daren't be too frank and open because they do compete. So their message to the vendor ends up being a series of individual requests with account teams, instead of one or two consolidated messages.
The competitive aspect also implies that, if carriers were more fairly represented in standards groups, that we'd see a similar in-fighting between them that we do when vendors deliberately take opposing views in the IETF today.
And as you rightly point out, carriers don't feel that they can change the way the IETF works, so they take their problem to a different standards group. But the ITU can't get involved in writing MPLS protocols, because that's the IETF's job. And if something deep inside the MPLS architecture needs to be changed to deliver good OAM (eg. getting rid of the PHP hack), then that's tough because the core protocols are already done, and submitted to the IESG.
So, who has a suggestion as to how this situation could be fixed?
re: The Chinese Are Coming, Says Poll>>"If we look at the goals of the MPLS Working Group and compare them to the original goals, we can see that the IETF has now hacked together MPLS into a working set of protocols that do actually allow connectivity to be achieved, but that additional MPLS protocols have yet to be developed to make this technology carrier grade."
Since when has IP been really carrier grade? Oops, I am really going to be flamed for that remark!
On the other hand, the carriers made their bed when they chose IP, so I guess they will have to lay in it since they could have chose ATM. Ouch! Another flame. BTW, IGÇÖm not against IP, itGÇÖs just a fact. And, maybe that is part of the problem; carriers keep comparing IP to carrier GÇôgrade TDM instead of envisioning it within IPGÇÖs best light.
Nevertheless, I respect the standards process because the stakes are so high, which makes politics a given. Moreover, I know quite a few engineers that write the IETF drafts, and although it may not look like it, they do listen to their customers and focus on getting the job done for them.
re: The Chinese Are Coming, Says PollHi Abby, No flames from me on those two statements :-)
Yes, you're talking to a fellow ATM and TDM fan. IMHO connections are essential for many carrier-grade features, and in particular for QoS and OAM. But I can see that ATM's day as the "convergence technology" is long gone. ATM will continue to hold strong in the ADSL and 3G markets, but elsewhere it's definitely lost the battle to IP.
But you see, I'm of the firm belief that we should learn from the best and worst in any technology - copy the good stuff and get rid of the bad. ATM and TDM experience should help us make MPLS better than it could have been if we only use IP origins.
I think the IP community has a bit of a problem admitting this - Scott Bradner and his supporters resisted MPLS for a long time on a point of detail (ie. in some applications MPLS packets may not contain an IP address, therefore the IETF has no business working on this technology).
There is still a faction in the IETF that refuses to make MPLS a special case, in effect refuses to recognise that the benefits brought by MPLS to IP may be the best chance to make IP into a carrier-grade technology.
My own feeling is that MPLS can never reach its full potential while this kind of philosophy is allowed to continue. It's not in the vendor interests to change the status quo in the IETF. It's really up to the victims of this situation - the carriers.
re: The Chinese Are Coming, Says PollOne thing that I have noticed is that all sorts of people want to discuss the sequence of bits in a standard but VERY few have any idea how to test what they have written. Often what happens is that the standard is written with a bunch of neat requirements and then many different test cases for that requirement appear from every different testing organization (vendors, carriers, and third-parties). If anyone has ever seen a Telcordia SONET compliance test report, all it lists are the requirements and 'Pass' or 'Fail'. They do not publish how the test was done. As some RBOCs still only accept their own, internal results, or a Telcordia test report, how is a vendor supposed to pre-test a system adequately for these trials?
What does everyone think of making technical submissions with an attached, detailed, test case so that there will be much less ambiguity when someone says that their product is 'compliant'? This would also imply that there would be 2 documents coming out of every standards activity: the actual standard and a separate, yet joined, test plan.
Your friends, the engineers who write IETF drafts, are probably doing their jobs as best they can. The carriers like to control as much as they possibly can for their own interests... sounds familiar somehow. I think what has gotten lost here is that the carriers don't necessarily want to make the technical submissions, they just want to have control over which submissions make it into the published standard.
Politics will always be involved as long as there are people involved, we can't change that. What would seem to make sense is only allowing carriers to vote on the content of the published standards. Allow everyone to make submissions, by all means, but give them (the carriers) the control that they will exercise in one way or another (eg: writing their own addendums, etc.) anyway. This might address Geoff's concern about carrier representation in these forums because they will see that the time spent is a means to an end and it will most likely save everyone a lot of time and effort.
re: The Chinese Are Coming, Says PollGeoff, You raise an interesting point - IETF deciding (based on single individual's POV) that an MPLS packet if it carriers non-IP payload, IETF wouldn't deal with transport/operational issues of such traffic.
It might be far fetched, but looks like either sub-IP group or new group should take responsibility to add to MPLS any missing pieces that you and others think will MPLS carrier grade technology - OAM is one for instance. It is always important to have peer fighting so any piece of idea gets refined upto some point -
It is also important in the event IETF refuses to undertake changes to MPLS, to take MPLS out-of IETF and put it in ITU/MPLSForum's control - why not - after all Awaduche and others started MPLS and MPLS doesn't belong to IETF (and if some body is suggesting IETF is owned by Cisco/Juniper then to these two companies). Carrier grade MPLS solutions if sourced as software only solution then vendors other than Cisco/Juniper can mushroom (there are few strong ones already). As long as customers/carriers have clear vision of what they need and can articulate them and get things standardized it will benefit industry as a whole.
If the carriers are not participating in a standard, for whatever reason, and therefore have no corporate ownership invested in that standard, how likely are they to make a corporate effort to implement and deploy that standard when they feel that the standard 'body' has reached its final version in the wrong way (from their point of view)?
re: The Chinese Are Coming, Says PollHi stephenpcooke, I totally agree with your post 97.
In the past the IETF has worked from the point of view of "running code" - in other words somebody throws a few lines of code together, lobs it out onto the Internet and sees what happens. By refining this running code, there is no need for irritating details like implementation agreements, PICS documents, test plans etc.
The original users of IP equipment could work within this kind of framework. But carriers cannot - or at least they would not if they had a choice.
It might be far fetched, but looks like either sub-IP group or new group should take responsibility to add to MPLS any missing pieces that you and others think will MPLS carrier grade technology - OAM is one for instance. It is always important to have peer fighting so any piece of idea gets refined upto some point -
Sub-IP looks a bit empty now that MPLS has been allowed into the Routing area :-)
I'm not sure how much my own opinion counts (or matters), but certainly I've spoken to lots of carriers in the past couple of years, and the focus of their MPLS concerns is definitely shifting to OAM now. Let's face, they're in the midst of deploying services like 2547bis and VPLS and right now the only way to manage them is either to send an engineer to hit the reset button on the switch, or to build an overlay control network so they can log into the console port.
It is also important in the event IETF refuses to undertake changes to MPLS, to take MPLS out-of IETF and put it in ITU/MPLSForum's control
A couple of years ago I would have heartily agreed with that idea. But you know, I don't think it'd do that much good.
It's clear to me that the IETF is close to the point where it has to face the problem of growing up. The fundamental processes of the IETF have to change.
In fact there's an interesting dicussion about how to deal with Internet Standards. If you're not on the mail list here is the link to the archive...
I would think the most optimistic way forward is for the IETF to accept that it has to change, and to make those changes as quickly and smoothly as possible.
Apart from the obvious procedural issues, such as the one highlighted in this thread, there are some very simple changes the IETF could make very easily...
- How about charging a membership fee? This would fund proper documentation and editing, for example.
- How about making it possible to put diagrams in an ID or RFC? The world has moved beyond ASCII, but the IETF has not :-)
It's tempting to say that the ITU or industry fora produce better standards than the IETF, and generally it's true. After all, the IETF never has to prove that a protocol works before it's published.
But when all's said and done, the IETF are the folks who own IP. Assuming they finally admit that MPLS is a new way of doing of IP then I guess that means they own MPLS too.
Been there, done that, and the T-shirt still smells of garlic :-)
Actually I worked in the vendor community for the past 20 years or so, and only became an analyst a couple of years ago.
It's much more fun for an analyst to look at the results of a standards group's works, than the way it achieves those results. That way, instead of getting down into the bits, bytes and territorial disputes, I can make sweeping observations like...
"If we look at the goals of the MPLS Working Group and compare them to the original goals, we can see that the IETF has now hacked together MPLS into a working set of protocols that do actually allow connectivity to be achieved, but that additional MPLS protocols have yet to be developed to make this technology carrier grade."
BTW I base this conclusion on the two current goals of the WG, which are OAM and inter-area operation. These are sine qua non for carriers.
I've had more experience of the Frame Relay Forum and ATM Forum approaches than the IETF. IETFs are great fun, but I'm not sure my humming is up to the massed choirs of Cisco and Juniper :-) However, I did follow the famous OAM thread a couple of years ago. If that's an example of how a WG works these days I wouldn't want to advertise it if I were you :-)
Only kidding. I think it would be far more beneficial for a carrier engineer to follow your advice, since they are the ones who have to try and get this disparate bag of protocols to work at the end of the day.
In contrast vendor engineers are focussed on achieving a competitive advantage for their existing technology. Have you seen one of Yakov Rekhter's latest conference presentations on why some vendors prefer to evolve protocols, while others prefer to develop new protocols?
Cheers,
Geoff