Microsoft has such a large influence on the desktop computing environment that any technology decision they make has the potential to swing adoption one way or another. Should we thank Microsoft for TCP/IP? I don’t think so: they merely chose it (after their own technically awful options failed to gain dominance) because it was freely implementable and did not aid their competitors.

Bear in mind that the Internet components in Windows 95 were not bundled as standard until the first service pack. In the original release, Microsoft still had hopes that their proprietary Microsoft Network could become a nice little walled garden, and support for this was integrated into the standard installation, whereas Internet support was only available through the “Microsoft Plus!” pack. They were a Johnny-come-lately to the whole “walled garden” network concept as well, where AOL was the dominant player (in the US market). Doubtless this was just another example of Microsoft hoping to gain market share in another area by using their Desktop OS monopoly as leverage.

Whatever the case, the Internet and the World Wide Web were becoming increasingly demanded at the time. Even prior to the release of Windows 95, people were installing dodgy Internet stacks and Netscape Navigator on Windows 3.1x. I think it’s fairer to say that Microsoft capitulated to the Internet and the Web, rather than having enabled its development. The decision to make Internet Explorer a core component in subsequent releases of Windows 95 was driven by the success of Netscape—a fact which was a key component in the antitrust trial. That capitulation may have hastened the demise of proprietary networking protocols and walled garden networks, but I doubt that it enabled it.

Love them or hate them, Microsoft is a convicted monopolist that can often make the industry go where they want it to go. We can occasionally breathe a sigh of relief when it bulldozes us all in a direction that we more-or-less wanted to go anyway. TCP/IP is one of those cases—Web standards, not so much.

Urm - reality check time.

Microsoft’s failure to bundle IP was a major delaying factor in deploying global IP easily.

Back in 1990 I graduated from a Univerity on the UK JANET which was an IP based network and joined an organization which already had an IP network (amongst other protocols), and there were two bug bears Microsoft and IBM.

IBM charged extra (a lot) for an IP stack on their mainframes at the time (I believe SAS use to sell an IP stack in competition for MVS), and Microsoft just didn’t have one (you had to install a third party stack).

But we managed to integrate these dinosaur systems into the network of Unix, Digital, and other open systems. Most of it was done at least half a decade before Windows 95 was released.

Around that time (must have been 1992) I remember saying “put documents in HTML” was an insightful response to “how will be best ensure we can share these documents electronically with other people in 5 years time, because we weren’t sure which Word Processor would still be around.

Okay that was an organization with strong academic links, but when I went out to industrial companies in 1995, everyone was already using IP for interoperability between systems in their data centers, if not further afield. The only folks not using IP were using Novell Netware, because they only needed to network one office, and several of the Netware based companies were running IP on Netware (that really did suck) for interoperability.

I don’t recall good old Trumpet Winsock being buggy. Although all that messing around because of the 640KB limit was a pain, but that wasn’t exactly their fault, 640KB was going to be enough remember.

You’ve spent too long in Redmond, and not long enough at the coal face. You’re beginning to sound like Al Gore.

Famous Brett, let me preface my comments by saying that I was the product manager at Microsoft responsible for IE, and the Windows 95 internet applications.  I was also one of the contributors to the company’s first web site, and one of a vocal group of people in the company who felt that the proprietary approach being pursued by the MSN team was destined to fail. 

Microsoft has such a large influence on the desktop computing environment that any technology decision they make has the potential to swing adoption one way or another. Should we thank Microsoft for TCP/IP? I don’t think so: they merely chose it (after their own technically awful options failed to gain dominance) because it was freely implementable and did not aid their competitors.

I didn’t say you had to thank the company.  I was merely observing that folks who are thanking Microsoft for “standardizing” the OS market missed the point.  As important as Windows 95 was to Microsoft, the importance of Windows 95 to the rest of the world was that it created a large, ubiquitous and vendor agnostic internet.  Because the market leader made it’s choice TCP/IP, the discussion over network protocols ended.

Bear in mind that the Internet components in Windows 95 were not bundled as standard until the first service pack. In the original release, Microsoft still had hopes that their proprietary Microsoft Network could become a nice little walled garden, and support for this was integrated into the standard installation, whereas Internet support was only available through the “Microsoft Plus!” pack. They were a Johnny-come-lately to the whole “walled garden” network concept as well, where AOL was the dominant player (in the US market). Doubtless this was just another example of Microsoft hoping to gain market share in another area by using their Desktop OS monopoly as leverage.

I completely agree that the company was late to the walled garden, and given the dominant position that others already had, it was unlikely they could succeed.  In a classic “hedge your bets” move, Gates and Allchin agreed to put both MSN and the Internet pieces into Windows 95.

It’s imprecise to say that the internet components weren’t bundled until the first service pack.  All versions of Windows 95 included the TCP/IP stack at launch.  The retail Windows 95 package did not include the browser, or the SMTP mail connector.  These were part of Windows 95 Plus!, a decision that I promoted to management because our research showed we could generate an extra $100 million in revenues through this packaging decision.  However, OEM versions of Windows 95 included the entire stack.  OEM windows at the time was selling at a rate of 50 million units annually and growing.  Upgrade windows (the retail version) was a relatively small business by comparison. 

Whatever the case, the Internet and the World Wide Web were becoming increasingly demanded at the time. Even prior to the release of Windows 95, people were installing dodgy Internet stacks and Netscape Navigator on Windows 3.1x. I think it’s fairer to say that Microsoft capitulated to the Internet and the Web, rather than having enabled its development.

“Capitulation” is an incorrect way to describe what happened.  The earliest instance of a TCP/IP stack being bundled standard with a Microsoft OS was the initial release of NT 3.1 in March of 1993.  This was in response to requests from corporate customers in the early 91/92 market. So in 1992 (during the development of NT) Microsoft was already on the TCP/IP bandwagon.

Mosaic itself didn’t release until August of 1993, and Navigator was released in October 1994.  At that point in time we already understood the desirability of an internet browser as a feature of the OS.  The Word team had already licensed BookLink and was including that as an add-on for Word in order to make it easy to author web pages and view them through Word.  The company was also negotiating to acquire BookLink when AOL swooped in and grabbed them.  That forced Microsoft to license Mosaic from Mosaic Communications in December of 1994. 

There was never a debate about whether to eventually integrate IE completely with the OS.  We could all see the value of having a single hypertext system for help, a single presentation system (the web) and so on.  You saw the first example of that in the incredibly awful active desktop which released with Windows 98.  It embedded HTML into the desktop, which turned out to be not very practical. 

Anyway, while Microsoft’s were Microsoft’s products and affected nobody but Microsoft’s customers, I stand by the assertion that Microsoft’s choice to end the network wars by enabling mass distribution of TCP/IP was by the far the most important choice of any that it made at that time.

Simon,

I wrote:

Microsoft’s decision to include IP as a native component in Windows 95 that was the company’s seminal contribution to today’s computing world. The company didn’t invent IP, didn’t own any intellectual property in TCP/IP, nor did it profit directly from it. However, by ensuring a relatively bug-free implementation of IP on the dominant operating system platform, Microsoft forced an open, standard and routable networking protocol on the world.

You won’t get any argument from me that TCP/IP existed before Microsoft, nor that there weren’t implementations on Microsoft platforms.  Trumpet was not for the faint of heart, but it did exist.  Frontier also had a complete stack. 

My argument is simply that Microsoft forced consistency, and that was a necessary pre-condition for the innovations that were to follow. 

Around that time (must have been 1992) I remember saying “put documents in HTML” was an insightful response to “how will be best ensure we can share these documents electronically with other people in 5 years time, because we weren’t sure which Word Processor would still be around.

Tim Berners-Lee’s first posting to the internet about the WWW project that he was working on at CERN was in August 1991.  It wasn’t until April 1993, though, that CERN would announce that the protocol would be free to use.  Up until that time the dominant internet application protocol was gopher.  Moreover HTML was woefully inadequate as a page description language at the time, no authoring tools existed, and browsers were incredibly crude.

It would have been very forward thinking to suggest HTML to your team. Perhaps you were suggesting SGML (common at the time), or TEX/LATEX?

I seem to remember Mosaic was fine.

But my point is Microsoft didn’t force consistency. The organizations I saw had consistency of networking protocol, and Microsoft was forced to play along with the other kids or be sidelined.

The Internet and the World Wide web already existed when Windows 95 arrived. Commercial Internet providers started in the UK in 1992, predating the web.

I don’t really see an awful lot of technical innovation in the web space. A lot of business and technical development, but most of the web looks a lot like NCSA httpd and Mosaic still. Does the same sort of thing. Probably the most dramatic shift is the DOM, and dynamic page content, but it is really a hack to work around the lack of a successful applet model.

The encryption I guess was a vital part of enabling ecommerce, but the implementations are pretty naff, and the protocol is poorly designed for its current use. Indeed the whole concept of a secure and an insecure channel is poorly presented to the user, I suspect a replacement could “just do secure” in these days of ample CPU.

What is more surprising I think it that the technologies built into the Internet have such huge inertia. SMTP should die. FTP really should have died a long time ago. HTTP is pretty solid, but the HTML and CSS standards are not exactly elegant and I’m surprised they haven’t been surpassed by newer approaches to the same problem that are just easier to use. HTTPS needs a revamp. DNS needs better security, but otherwise I think it does its job reasonably well.

Google and Skype I think stand out as examples of technical innovation. Or for Skype at least bringing together diverse technologies into one easy to use application, although it largely enabled something that already existed (telephony). Instant messaging has come on a long way.

But of course the big thing about Skype was the speed of arrival. We have the technology to deploy a “clearly better” or innovative solution to all these problems to everyone on the net in a matter of days or hours. All we need is people to figure out what they are and code them.

I suspect Skype’s success was largely down to the lack of need for administrative involvement. The reason SMTP hasn’t been replaced may be the (perceived?) need for managed servers that hold your email when you aren’t online.

This seems like highly revisionist history from a pro-Microsoft bias.  The way I recall it, the Internet and the World Wide Web were taking off like wildfire in the early months of 1995 (after several years of being used mostly in academia), and in addition to direct ISP access via things like Trumpet Winsock, the Web and other Internet features were starting to become available within proprietary online services like Prodigy (which launched its own Web browser in Feb. ‘95), CompuServe, and AOL.  The Web was becoming a juggernaut that would have become ubiquitous regardless of what Microsoft did or didn’t do, and MS was actually pretty late jumping on that bandwagon.

Alec Saunders said:

I was the product manager at Microsoft responsible for IE, and the Windows 95 internet applications.  I was also one of the contributors to the company’s first web site, and one of a vocal group of people in the company who felt that the proprietary approach being pursued by the MSN team was destined to fail.

Perhaps, then, you could tell why IE invariably ignores compliance to standards. One thing I grasped, for example, is that IE’s behavior has been adjusted so as to compensate a lack of IIS features, neglecting to consider what happens when the content is served by other software.

Email, the other relevant Internet application nowadays, has seen implementations notoriously entangled with MAPI. Microsoft’s role in the MARID case is another example of standardization difficulties. I don’t mention the recent standardization of MS-Office, since documents formats have already been discussed.

As for the TCP/IP stack, Windows APIs still include socket() along with WSASocket(), consistently easing portability in a way aimed at forcing a not so open MS-standard protocol on the world.

Microsoft has a mighty programming power. However, its ability to bring consistency in a useful way is hindered by allowing marketing considerations to influence technical developments.

Ale,

IE’s continued lack of standards compliance is frustrating to everyone, myself included.  My personal blog has style sheets for IE 6, IE 7, and Firefox because of it.  However, it’s been over 10 years since I had anything to do with IE, and I left the company in 2000, so I really can’t answer your questions about Microsoft’s current approach to standards.