Yusuke Aoyama
IPv6style
IPv6style discussed IPv6 support by Windows Vista many times. We run articles about not only Windows Vista itself, but influence of the spread of Windows Vista to networks.
But we haven't really talked about the fundamental issue of why Windows Vista had to support IPv6 by default.
Therefore, we interviewed Takuya Oikawa, Group Program Manager, Windows Platform Group, Windows Development at Microsoft Development, Ltd., the reason Windows Vista supports IPv6 and future computing vision by Microsoft.
[0]
Reasons Vista is given IPv6 support by default
IPv6style: Why is IPv6 necessary for Windows Vista?
Oikawa: Original direction of Windows and Microsoft software is seamless computing.
Seamless computing means computing environment with always-on connection anytime and anywhere, enabling various networked applications to run.
In order to realize it, we definitely need global reachability. Traditionally with IPv4, global reachability has been hampered by firewalls and NATs.
Therefore, Vista aims to enable seamless, anytime-and-anywhere network environment, utilizing global reachability of IPv6.
IPv6style: Do you mean that seamless computing envisioned by Microsoft is difficult to realize with traditional IPv4?
Oikawa: That's right. There have been a lot of efforts to prolong the life of IPv4, and some of such efforts have been paid off. But, to put it bluntly, these measures have been patchworks.
Technology such as UPnP tries to enable peer-to-peer, end-to-end communications. But such technology cannot be a perfect solution, as they incur some inevitable costs here and there.
We will see more and more non-PC devices to connect to the Internet. When such multiple devices receive services simultaneously, UPnP and other solutions do not work in a perfect manner. We consider that IPv6 is appropriate for fundamental solution.
IPv6style: So that's the reason for default IPv6 support in Vista. But IPv4 is the predominant protocol in reality. Isn't it a little to early for IPv6 to be included in the basic spec of an OS?
Oikawa: It is not particularly too early. Lifecycle of an operating system can span several years, or nearly 10 years in some cases. Therefore, we consider that complete transition to IPv6 will occur in the lifecycle of Vista.
As a matter of course, some customers move fast and others don’t. But early customer migrations to IPv6 will begin in several years. That's why Vista needs to support IPv6.
IPv6style: With Windows XP, IPv4 cannot be turned off to leave IPv6 working alone. On the contrary with Vista, IPv6 cannot be turned off to leave IPv4 working alone. How can we deal with the networking problems we may encounter with IPv6?”
Oikawa: IPv6 proponents might feel discouraged, but IPv6 can be turned off in fact. It's just that IPv6 uninstall is not as easy as IPv4 uninstall.
You can do a complete uninstall of IPv4 at command line using “netshell interface ipv4 uninstall”, while “IPv6 uninstall” command is not available.
However, Windows has a feature called group policy, which enables central control of clients in a domain. IPv6 can be disabled in this group policy setting.
IPv6style: Is that offered as a preparation for possible troubles on Intranet?
Oikawa: This method is not something that we recommend. However as you say, there is a possibility of some problems of misbehaviors of Ethernet switches, routers, network-attached printers and copy machines when they receive IPv6 packets.
When these problems happen, we want the vendors of the devices to fix them. We made it possible to put IPv6 off temporarily till they can cope with the issues.
[0]
IPv6 is a natural direction for Microsoft
IPv6style: I imagine there were a lot of discussions in Microsoft in giving a new OS default support of IPv6. What was the actual situation?
Oikawa: I asked the headquarters in Seattle about that, but they told me there were not so much of discussions.
Ipv6style: Was it a straight forward decision to go with IPv6?
Oikawa: That's right. When we develop an operating system, we set out some rules first. Windows Vista's code name is Longhorn, so the rule for the OS is called “Longhorn Basics”.
We build huge OS code base, so the basics dictate what principles we need to abide by in writing each source code, as a sort of creed. One of the principles in it is IPv6 support.
Longhorn Basics has 10 to 20 items in total. It can be pointed out on one letter-size sheet. If any important issue arises concerning any of the Basic principles, it needs to be fixed even at the very last stage of development without fail.
We haven't made it public, but there is no problem if it is made open. The principles laid out are very general ones in developing a new OS, highlighting which items needs to be focused on.
Principles other than IPv6 support includes “single binary support”, common binary across worldwide versions, which we realize for the first time. The Basics further includes Unicode support, a necessary part of common binary. IPv6 support is laid out as a principle just as these are.
IPv6style: Who first proposed IPv6 support to be included in Longhorn Basics? And what process did the proposal go through for approval?
Oikawa: Networking team in the headquarters. Networking team said they wanted to include IPv6 support themselves.
The reason is based on the needs by worldwide national institutions and governments, as well as networking vendors and carriers. As I said, IPv6 is expected to be wide spread during the lifecycle of Vista, so support was absolutely necessary.
Bill Gates himself went over the Longhorn Basics many times during the approval process. IPv6 support was included on complete understanding by the top management team.
IPv6style: Who actually developed IPv6 stacks?
Oikawa: As with Windows Vista, the features were mainly developed by development team at Redmond headquarters. But the Japanese were actively involved in the verification process. In fact, no other countries in the world have commercial IPv6 services which allow feature testing at commercial level. Therefore, Japanese subsidiary has made considerable contribution in verification and spec design.
IPv6style: As for verification, Vista is in public beta. How many feedbacks have you received?
I can’t tell you the number of comments we received, but we got considerable feedbacks. In Japan, we have some partners involved in joint-development-like efforts. One of them is very aggressively promoting IPv6. We are closely cooperating with the company for verification.
We have asked several partners to confirm that our features are satisfactory for all service products they provide, as well as share some of our future plans.
We get many feedbacks about IPv6 from general technical beta users, but they cannot cover the full range of issues. Therefore, we cooperate closely with partners in verifying our features.
What Microsoft can do for the spread of IPv6IPv6style: You just said that Japan is the only country with commercial IPv6 services widely available. In fact IPv6 is available in few countries even if users may want to use it. What can Microsoft do to change such situation?
Oikawa: Education is the main thing. One thing to do is to tell the world that technology for transition is established and available.
Many people believe that they can’t use IPv6 unless carriers and service providers offer IPv6 services, but that’s not true.
Microsoft and others offer transitioning technology. So we need them to fully realize that IPv6 can be used on IPv4 networks.
The other thing is to ask application developers develop programs independent of IPv4. We offer protocol-independent API for that.
We are telling application developers, “please fix your code so that it can be used on IPv4 and IPv6, even if you might use the program only on IPv4. Then you can derive the advantages of IPv6 immediately when IPv6 becomes available in your countries”.
IPv6style: I see. How many of current Windows applications can work on IPv6 as is when they are moved to Vista?
Oikawa: To tell you the truth, there are not so many. Applications using sockets tend to have address space variables and structure dependent on IPv4. Many applications need fixes.
But to go back to the topic of no disagreements within Microsoft, the reason we had no major disagreements is that Microsoft applications were easy to migrate.
Applications using sockets are difficult to migrate, but not many Microsoft applications directly utilize sockets. Many of them use higher level APIs.
For example, applications using HTTP and browser control are not directly aware of IP addresses.
Many applications with network support which do not focus on networking use higher level APIs, so they can work on IPv6 as is. That’s one reason IPv6 met few objections in Microsoft.
IPv6style: How do you confirm IPv6 support by applications?
Oikawa: We have a migration tool called checkv4. You input your code and it tells you which codes are IPv4 dependent and how to fix them.
We have a separate in-house IPv6 support check tool. Source code checking cannot be done unless the codes first go through this tool. This particular tool not only checks IPv6 support but also buffer overflow, for example. So buffer overflow and IPv6 support are checked at the same level. Unless you fix the code as the tool tells you, your Vista source code cannot be checked in.
We also have a checklist for internal use. This list confirms what tests have been conducted on each component. The list includes testing on pure IPv6 environment. So without IPv6 testing, the particular component cannot be approved for release.
IPv6stlye: You talked about activities to direct Microsoft customers such as developers and end users. Are there any other activities Microsoft can do to help spread IPv6?
Oikawa: We maintain close collaborations with partners. In Japan, they include networking device vendors, carriers and service providers. We actively communicate with them.
Core IPv6 specs were fixed, but additional standardization is made here and there. Also, Microsoft cannot decide alone what feature to be implemented when. We ask opinions to our partners on them.
Networking products cannot necessarily get connected with one another just because they are built on spec, or RFCs. Therefore, we actively participate in Moonv6 and other interoperability testing event to verify interoperability with many other vendors.
Advantage of peer-to-peer envisioned by Microsoft
IPv6style: What do you think is the most major advantage of IPv6 that end users enjoy?
Oikawa: At the risk of being repetitious, it’s the reachability realized by the absence of NATs. In other words, perfect peer-to-peer environment.
Current peer-to-peer applications implement various mechanisms under the hood. If one method fails, then the app tries another. If that fails, then the app tries another one. Some of them use servers for communications while they say they are peer-to-peer.
Various efforts are made in realizing peer-to-peer. By moving to IPv6, we can lower the development cost for these applications.
IPv6style: Could you be more specific about that?
Oikawa: Let’s say several people a conduct conference here at Microsoft, for example. Non-Microsoft people unfortunately cannot get connected to our internal network for mail and file exchanges or share presentations.
But with Vista, we can build an ad-hoc network just for this purpose. For Example, IEEE 802.11 wireless LAN ad-hoc connection can be enabled on each PC to create a network, and share files or they can see my presentations on their own PCs page by page as I conduct one.
Vista is standard equipped with some applications that allow s collaborative work by all people connected on a specific file without a projector.
IPv6style: You wrote in the blog by Microsoft development that peer-to-peer is required to support real-time communications, collaborations, contents distribution, and distributed processing. I’m aware that Vista has applications to cover the first three factors. What about distributed processing?
Oikawa: We have no applications we can specifically talk about just yet. There are many plans for grid computing like applications, but we need to consider which works well on IPv6.
Distributed computing or other topics, Microsoft has two roles. One is to supply all products required for a particular solution, as a software vendor.
But on the other hand, we are a platform vendor. We would like our partners and users themselves to build applications on our platform.
In Vista, most components support IPv6. Many applications of ours other than Windows increasingly support IPv6. But there are several aspects where we fall short of. We want other companies and partners to offer such solutions.
Of the four points I raised in the blog, distributed computing is indeed one area that we have not given specific form of solution. We can people with good ideas to build such system on Windows.
IPv6style: Grid Computing appears to be attracting more attention as one of the solutions for distributed processing. How helpful do you think the technology will become?
Oikawa: This technology will be very helpful. But users and the society need to get ready for it. Should such computing be offered for free or for money? There should be a lot of thinking to do on the business model.
For example, we need to start by considering the mechanism of how I can be paid for temporarily offering part of my resources. Many issues need to be clarified in order to build a successful model.
In addition, people are used to offer their own resources to others if temporarily only in one secure domain, such as one company. Projects like SETI@home have been available, but that’s really a client-server model conducting batch processing.
Many obstacles need to be overcome before this technology is widely deployed in a true sense. We need to solve issues on security as well as business. So we need some time.
IPv6style: Will Grid Computing be easily utilized by end users?
Oikawa: Before becoming a personal technology, it may bring easy-to-use mechanisms to high-performance computing.
IPv6style: Thank you very much.