Nathan Evans' Nemesis of the Moment

WebSocket servers on Windows Server

Posted in .NET Framework, Software Design, Windows Environment by Nathan B. Evans on December 20, 2011

This is a slight continuation of the previous WebSockets versus REST… fight! post.

Buoyed with enthusiasm of WebSockets, I set about implementing a simple test harness of a WebSockets server in C#.NET using System.Net.HttpListener. Unfortunately, things did not go well. It turns out that HttpListener (and indeed, the underlying HTTP Server API a.k.a. http.sys) cannot be used at all to develop a WebSockets server on current versions of Windows. The http.sys is simply too strict with its policing of what it believes to be correct HTTP protocol.

In an IETF discussion thread, a Microsoft fellow called Stefen Shackow was quoted as saying the following:

The current technical issue for our stack is that the low-level Windows HTTP driver that handles incoming HTTP request (http.sys) does not recognize the current websockets format as having a valid entity body. As you have noted, the lack of a content length header means that http.sys does not make the nonce bytes in the entity available to any upstack callers. That’s part of the work we will need to do to build websockets support into http.sys. Basically we need to tweak http.sys to recognize what is really a non-HTTP request, as an HTTP request.

Implementation-wise this boils down to how strictly a server-side HTTP listener interprets incoming requests as HTTP. For example a server stack that instead treats port 80 as a TCP/IP socket as opposed to an HTTP endpoint can readily do whatever it wants with the websockets initiation request.

For our server-side HTTP stack we do plan to make the necessary changes to support websockets since we want IIS and ASP.NET to handle websockets workloads in the future. We have folks keeping an eye on the websockets spec as it progresses and we do plan to make whatever changes are necessary in the future.

This is a damn shame. As it stands right now, Server 2008/R2 boxes cannot host WebSockets. At least, not whilst sharing ports 80 and 443 with IIS web server. Because, sure, you could always write your WebSocket server to bind to those ports with a raw TCP socket and rewrite a ton of boilerplate HTTP code that http.sys can already do, and then put up with the fact that you can’t share the port with IIS on the same box. This is something that most people, me included, do not want to do.

Obviously it isn’t really anyone’s fault because back in the development time frame of Windows 7 and Server 2008/R2 (between 2006 to 2009) they could not have foreseen the WebSockets standard and the impact it might have on the design of APIs for HTTP servers.

The good thing is that Windows 8 Developer Preview seems to have this covered. According to the MSDN documentation, the HTTP Server API’s HttpSendHttpResponse function supports a new special flag called HTTP_SEND_RESPONSE_FLAG_OPAQUE that seems to suggest it will put the HTTP session into a sort of “dumb mode” whereby you can pass-thru pretty much whatever you want and http.sys won’t interfere:

HTTP_SEND_RESPONSE_FLAG_OPAQUE

Specifies that the request/response is not HTTP compliant and all subsequent bytes should be treated as entity-body. Applications specify this flag when it is accepting a Web Socket upgrade request and informing HTTP.sys to treat the connection data as opaque data.

This flag is only allowed when the StatusCode member of pHttpResponse is 101, switching protocols. HttpSendHttpResponse returns ERROR_INVALID_PARAMETER for all other HTTP response types if this flag is used.

Windows Developer Preview and later: This flag is supported.

Aside from the new System.Net.WebSockets namespace in .NET 4.5, there are also clear indications of this behaviour being exposed in the HttpListener of .NET 4.5 through a new HttpListenerContext.AcceptWebSocketAsync() method. The preliminary documentation seems to suggest that this method will support 2008 R2 and Windows 7. But this is almost certainly a misprint because I have inspected these areas of the .NET 4.5 libraries using Reflector and it is very clear that this is not the case:

The HttpListenerContext.AcceptWebSocketAsync() method directly calls into System.Net.WebSockets.WebSocketHelper (static class) which has a corresponding AcceptWebSocketAsync() method of its own. This method will then call a sanity check method tellingly named EnsureHttpSysSupportsWebSockets() which evaluates an expression containing the words “ComNetOS.IsWin8orLater“. I need say no more.

It seems clear now that Microsoft has chosen not to back port this minor HTTP Server API improvement to Server 2008 R2 / Windows 7. So now we must all hope that Windows Server 8 runs a beta program in tandem with the Windows 8 client, and launches within a month of each other. Otherwise Metro app development possibilities are going to be severely limited whilst we all wait for the Windows Server product to host our WebSocket server applications! Even still, it’s a shame that Server 2008 R2 won’t ever be able to host WebSockets.

It will be interesting if Willie Tarreau (of HA-Proxy fame) will come up with some enhancements in his project that might benefit those determined enough to still want to host (albeit, raw TCP-based) WebSockets on Server 2008 R2.

Advertisements
Tagged with: , ,

WebSockets versus REST… fight!

Posted in .NET Framework, Software Design by Nathan B. Evans on December 16, 2011

Where will WebSockets be on this InfoQ chart, in three years time?

On 8th December 2011, a little known (but growing in awareness) standard called “WebSockets” was upgraded to W3C Candidate Recommendation status. That is one small step shy of becoming a fully ratified web standard. And just to remove any remaining possible doubt: Next-gen platforms and frameworks such as Windows 8 and .NET 4.5 (at three levels: System.Net, WCF and ASP.NET) already have deeply nested support, and they aren’t even beta  yet!

After reading up about the standard in detail and absorbing the various online discussion around it, it is becoming increasingly clear to me that this standard is going to steal a large chunk of mind share from RESTful web services. What I mean is that there will come a stage in product development where somebody will have to ask the question:

Right guys, shall we use WebSockets or REST for this project?

I expect that WebSockets will, within a year or two, begin stunting the growth of RESTful web services – at least as we know them today.

What are WebSockets?

They are an overdue and rather elegant protocol extension for HTTP 1.1 that allows what is fundamentally a bi-directional TCP data stream to be tunnelled over a HTTP session. They provide a built-in implementation of TCP message framing, so developers don’t need to worry about any boilerplate code stuff like that when designing their application protocol.

Why are WebSockets a threat to RESTful web services?

From the last few years of working on projects that expose RESTful web services, I have noticed a few shortcomings. I should probably make clear that I’m not claiming that WebSockets answers all those shortcomings. I’m merely suggesting that REST is not the silver bullet solution that it is often hyped up to be. What I am saying is that there is definitely space for another player that can still operate at “web scale”. WebSockets have more scope to be a little more like a black box or quick’n’dirty solution than REST which requires a more design up-front approach due to versioning and public visibility concerns. Always use the correct tool for the job, as they say.

Sub-par frameworks

They might claim REST support but they still haven’t truly “groked” it yet, in my opinion. WCF REST is a good example of this. Admittedly, the WCF Web API for .NET is starting to get close to where things should be, but it is not yet production ready.

Perhaps even more serious is the lack of widespread cross-platform RESTful clients that work in the way that Roy prescribed; of presenting an entry point resource that allows the client to automatically discover and autonomously navigate between further nested resources in a sort of state machine fashion. A single client framework that can operate with hundreds of totally different RESTful web services from different organisations. This does not exist yet, today. This is why so many big providers of RESTful web services end up seeding their own open source projects in various programming languages to provide the essential REST client.

Enterprise loves SOAP (and other RPCs)

Third-parties that want to use your web services often prefer SOAP over REST. Many haven’t even heard of REST! WebSockets are a message-based protocol allowing for SOAP-like RPC protocols that enterprise seem to adore so much. Hell there’s nothing stopping actual SOAP envelopes being transferred over a WebSocket!

This might not be the case if you’re operating in an extremely leading edge market such as perhaps cloud computing where everyone is speaking the same awesomesauce.

Complex domain models

Mapping out complex domain models onto REST can be slow and labourious. You’ll find yourself constantly having to work around its architectural constraints. Transactions, for example, can be a particular problem. Of course, this is partly related to the first problem (sub-par frameworks) but one cannot reasonably expect transaction support in a REST framework. What is probably needed is a set of common design patterns for mapping domain models to REST. And then an extension library for the framework that provides reusable implementations of those patterns. But alas, none of this exists yet.

Text-based formats

JSON/XML (for reasons unknown) are commonly used with REST and these are of course text-based formats. This is great for interoperability and cross-platform characteristics. But it is not so great for memory and bandwidth usage. This especially has implications on mobile devices.

You’ll find yourself running into walls if you try to use something that isn’t JSON or XML, at least that is my experience with current frameworks.

Request-response architecture of HTTP

Fundamentally, REST is nothing more than a re-specification of the way HTTP works and a proposal of a design pattern to build applications on top of HTTP. This means it retains the same statelessness and sessionless characteristics of HTTP. It therefore precludes REST from being bi-directional where the server could act as the requester of some resource from the client, or sender of some message to the client. As a result it requires “hacks” to be used to emulate server-side events, and these hacks have bad characteristics such as high latency (round trip time) and are wasteful of battery life.

Public visibility, versioning concerns

Sometimes having everything publicly visible is not what you want. People start using APIs that you don’t want them to use yet. You have to design everything to the nth degree much more. Have a proper versioning strategy in place. It encourages a more discerning approach to software development, that is for sure. Whilst these are usually good things, they can be a hindrance on early stage “lean agile” projects.

What can WebSockets do that is so amazing?

The fact that there will soon be a second player in this space suggests that there will be rebalancing of use-cases. WebSockets will prove to be disruptive for several reasons:

True bi-directional capability and server-side events, no hacks

Comet, push technology, long-polling etc in web apps are slow, inefficient, inelegant and have a higher potential magnitude for unreliability. They often work by requesting a resource from the server, causing the server to block until such a time that an event (or events) need to be transferred back to the client. They can be unreliable because the TCP connection could be teared down by a intermediate router during the time it is waiting for the response. Or worse, a proxy server might deliberately  time out the long-running request. As such, many implementations of this hack will use some kind of self-timeout mechanism so that perhaps every 60 seconds they will reissue the request to the server anyway. This has implications on both bandwidth and battery usage.

The true bi-directional capability offered by WebSockets is a first for any HTTP-borne protocol. It is something that neither SOAP nor REST have. And which Comet/push/long-polling can only emulate, inefficiently. The bi-directional capability is inherently so good that you could tunnel a real-time TCP protocol such as Remote Desktop or VNC over a WebSocket, if you wanted.

Great firewall penetration characteristics

WebSockets can tunnel out of heavily firewalled or proxied environments far easier than many other RPC designs. I’m sure I’m not alone in observing that enterprise environments rarely operate their SOAP services on port 80 or 443.

If you can access the web on port 80 without a proxy, WebSockets will work.

If you can access the web on port 80 with a proxy, WebSockets should work as long as the proxy software isn’t in the 1% that are broken and incompatible.

If  you can access the web on port 443 with or without a proxy, WebSockets will work.

I strongly suspect that there will be a whole raft of new Logmein/Remote Desktop and VPN solutions that are built on top of WebSockets, purely because of the great tunnelling characteristics.

Lightweight application protocols and TCP tunnelling

There is the potential for extremely lightweight application protocols, in respect of performance, bandwidth and battery usage. Like REST, the application schema/protocol isn’t defined by the standard; it is left completely wide open. WebSockets can transfer either text strings or binary data. It is clear that the text string support was included to aid in transferring JSON messages to JavaScript engines which lack the concept of byte arrays. Whilst the binary support will be most useful tunnelling TCP streams or for custom RPC implementations. After a WebSocket session is established, the overhead per message can be as small as just two bytes (!). Compare that to REST which has a huge HTTP header to attach to every single request and response.

How will the use-cases of REST change?

I believe that REST will lose a certain degree of its lustre. Project teams will less eagerly adopt it if they can get away with a bare bones WebSocket implementation. REST will probably remain the default choice for projects that need highly visible and cross-platform interoperable web services.

Projects without those requirements will probably opt for WebSockets instead and either run JSON over it, or use a bespoke wire protocol. They will particularly be used by web and mobile applications for their back-end communications i.e. data retrieval and push-events. Windows 8 “Metro” applications will need to use them extensively.

I suppose you could summarise that as:

  • REST will be (and remain) popular for publicly visible interfaces.
  • WebSockets will be popular for private, internal or “limited eyes only” interfaces.

Note: By “public” and “private” I am not referring literally to some form of paid/subscription/membership web service. I am referring to the programming API contract and its level of exposure to eyes outside of your development team or company.

Conclusion

Even though they are competing, the good thing is that REST and WebSockets can actually co-exist with one another. In fact, because they are both built upon HTTP fundamentals they will actually complement each other. A RESTful hypermedia resource could actually refer to a WebSocket as though it were another resource through a ws:// URL. This will pave the way for new RESTful design patterns and framework features. It will allow REST to remedy some of its shortcomings, such as with transaction support; because a WebSocket session could act as the transactional unit of work.

The next year is going to be very interesting on the web.