2003: Year of Apache

John Chamberlain writes "Netcraft's numbers for the new year are in. The trend graphs tell a story: 2003 was the Year of Apache. If Time magazine had a server-of-the-year award the cover would be featuring a feather. Since October 2002 market share has grown from 53% to 64%, a 20% gain while Microsoft IIS, its nearest competitor has shrunk from 36% to 24%, a 33% decline. The change in server totals was even more dramatic. Apache HTTP Server increased from about 20 million to 32 million (+60%) while all other competitors remained flat."

A win for open source

[mgv]

The big advantage of measuring the fall in IIS vs Apache is that web servers are public, and easily counted.

I'm sure that the same thing is happening thoughout the open source movement, but its just alot harder to measure the number of (for example) Linux installs when there is no central body that really collects data on this (not that there is any need for this).

So its representing a victory for much more than Apache.

Michael

Re:20 percent gain?

[Sp4c3 C4d3t]

But it also happens to be about 20% more than 53... I think that's what they mean.

TCO

[Nadsat]

MS's recent campagin of Total Cost Of Ownership does not factor well into this. They cite recent studies which heavily stress human maintenance and development costs into the TCO. Yet what they don't cite is the fact that as software popularity grows, such as Apache here, TCO is driven down because the technology is more accesible.

Basic technology such as web servers are on their way of being removed from the realm of competition. 2004 is promising.

Re:The platform they did the calculations on

[benja]

A 33% decline is a decline by a third. A decline from 36% to 24% is also a decline by a third. Ok, there was rounding involved. What's your problem?

Build a better mousetrap...

[sdo1]

Is anyone surprised? It's a superior piece of software from the competition. And the users (meaning IT folks and people who run web sites) are not your average Joe Blow, so having open source software makes absoulute sense. It's not like a desktop app (like a word processor) where the person using it would have a need or want or ability to go mess with the code for some reason.

Additionally, any serious security bugs have been fixed with blazing speed. Compare that with the amount of time MS takes to patch a IIS hole when an exploit is found.

-S

Yes, but measuring webserver market share is hard

[Raul654]

Web server market share is a funny thing. Do you count the total number of webservers, or just domains? What if you use a very ineffecient implimentation, and it takes twice the number of machines to do it? Should the server get a better market share because of it? The numbers are open to a lot of intepritation.

Apache 2 runs well on Windows

[hey]

If you assumed Apache was *nix only you haven't checked out Apache 2.x on Windows. Perhaps this is the cause of the gain -- Windows users switching to Apache?

Re:Makes you wonder

[Stile 65]

Actually, the post says the number of servers for IIS stayed flat. Their percentage decreased, but that seems to be a function of a huge number of ADDITIONAL web servers, of which an enormous percentage are Apache rather than IIS.

"across all domains" ?

[ubiquitin]

Note that the numbers are "per domain" So 2003 is better proclaimed the Year of NameVirtualHost. Hopefully, this means that there really are more httpd's out there, but the correlation was not made in that necraft study. Hopefully someone will do (perhaps already has done?) a study to establish IP# to domain name ratios. My guess is that there is a lot more virtual hosting being done now then there was in, say 1999, when having a corporate web site was more directly related to purchasing dedicated web server equipment. I'll bet that the Microsoft push into public key infrastructure will be used to leverage growth for IIS but at these rates, it may well be hard to catch up with Apache.

But perhaps the real story for 2003, as far as growth technologies go, is likely PHP [php.net]. The ratio of deployments and actual usage to press coverage of the technology is pretty impressive too. :)

Re:Yes, but measuring webserver market share is ha

[rsheridan6]

Shouldn't these issues remain fairly constant? Maybe it's tricky to count market share in absolute terms, but the trend-line should be pretty accurate.

Re:Who *are* these guys?

[maelstrom]

You make a very good point. Many people can list off Larry Wall, ESR, RMS, and Linus off the top of their head, but don't know the first thing about the principles in the Apache project. Seems to be a nice counter-point to ESR's ego currency as a motivation for OSS. Apache is in my mind the most successful OSS project.

Kudos.

Re:I wonder, why...

[Tom]

"because it works"

I've run apache on all kinds of systems, from the older pentiums you mention to big-iron Solaris systems.

The beauty is that it works on all of them. You tune some parameters slightly different, but you don't have to learn a new software because you're now hosting your site on a big machine.

Sorry, I applaud all the tiny-http-server efforts, but in real-life the only thing that I ever seriously considered was the kernel-httpd. That was for the image-server of a major dot-com site that made a several hundred hits a second at peak times.

Re:Mac OS?

[Anonymous Coward]

Being that MacOSX's "personal web sharing" *is* Apache, could be that an OS with single digit marketshare may have an impact. Doesn't need to be a server OS to be part of the mix.

Re:Makes you wonder

[ClosedSource]

So, the question for MS is what percentage of those additional servers were operated by potential IIS customers as opposed to individuals or organizations that simply wouldn't operate a site if the server wasn't free?

It's bit like the complaints from the record companies about how much money they lost to illegal downloading: the downloaders couldn't possibly afford to pay for all the music they download, so the actual losses are a lot smaller.

Viruses, Worms, and Exploits Are... Where?

[the_mad_poster]

Alright - let's have it! Where are they hiding all the exploits? They obviously have waaaayyyyy more since viruses and exploits are dependant on popularity, not how well the software is engineered. Since Apache is kicking IIS's scraggly ass all over the 'net, it must have more exploits, right? No? Oh? So all those people that keep saying Windows suffers so much are admitting they're wrong?

Oh, that's right. IIS is also an FTP server, mail server, dinner server, and a cheauffer that takes your wife out on dates then screws her in your bed while you're out of town on business.....

... whoops.. sorry, go a little carried away there. Seriously - face it, that's a flaw. If the software wants to do everything, and, by doing everything, fails, it still failed, AND it failed BECAUSE it does everything. That means the Apache software is a better engineered web server and IIS is, well, a load of crap.

Sorry... a little bitter. If you've ever had to administer that horrendous piece of garbage IIS you'd understand. I think, perhaps, the reason Apache is whooping up on IIS is that IIS is so ludicrously twitchy and convoluted. Normally, I'd say point and clicky interfaces are easier to manage, but god... setting something up in IIS that's not set up by default can result in tremendously time-wasting efforts searching through numerous, poorly labeled, badly designed interfaces. Apache? Whip out a reference book, type in a few lines, and you're done. Even if you have to restart the system, it's not much hassle. I've NEVER managed to shut down IIS and bring it back up on Win2k where it didn't stop responding and, eventually, chew up all the resources on the box forcing a hard reboot of the whole system. That pisses off SQL Server which then fucks up the TrendMicro stuff... Ick.

Long story short? IIS sucks and few (smart) people debate that whether they're pro-Microsoft, pro-*nix, pro-Mac, or, smarter than any of them pro-whatever-works.

Re:Yes, but measuring webserver market share is ha

[sosume]

It's even funnier, since Netcraft only counts public webservers. They do not include the zillion corporate intranet servers that used to be publicly available shielde by only NTLM authentication.
Thanks to the blaster outbreaks and the growing number of vpns these servers are now shielded off the regular internet. And thus the number of IIS in Netcraft's reports declines..

Re:Apache is racist.

[Trurl's Machine]

OK, I see it's a troll, but... I'll bite anyway. The whole history of Western Civilization in a nutshell looks like that. Once there were the pre-hellenic mediterranean cultures, like the Phoenicians. The Greeks conquered and destroyed them all. Then came the Romans, who conquered and destroyed the Greeks (not to mention the Celts). Then came the German and Slavonic barbarians, and they conquered and destroyed the Romans (and then repeatedly conquered and destroyed each other, like the Goth who perished for the Vandals etc). So if you live, say, in London, there are ashes of dozens of destroyed cultures under your feet, under the pavement of the very Oxford Street. The Celts, the Celtic-Romans, the Roman-Romans, the Anglo-Saxons, the Anglo-Danes, the Normans etc.

Now, for a long time Americans were fed with the not-exactly-true fairy tale about the Mayflower settlers, who arrived to a no-man's-land. It was not a no-man's-land at all. It had its native inhabitants and they were, indeed, conquered. But the British Islands were not a no-man's-land neither, when the William the Conqueor arrived, and he is still regarded as hero.

There's nothing racist in Apache, just as there's nothing racist when modern Britons use greek, latin, saxon or celtic words. Or when modern Italians use the name La Fenice (="country of Phoenicians") for an opera.

Sorry for feeding trolls.

Re:Yes, but measuring webserver market share is ha

[Nexx]

No, that's an indicator of server quality for that purpose. If the majority of server operators didn't want virtual hosting, for example, IIS not playing well in that environment won't make a shred of a difference.

These surveys also do not count the millions of intranet-only sites that these servers serve, and given the nature of the beast, I'm going to guess IIS is rather prevalent in that market.

I have recommended IIS-based solutions before, and given the same requirements, I'll do it again.

Bzzzzt. Thanks for playing

[Anonymous Coward]

I think you've just illustrated the mans point.

You've listed 4 Apache installations. Big deal.

> But I'm certain that for every half-assed amateur using
> Apache there are 100 admins who run Apache for
> mission-critical stuff and don't bat an eyelid

Would you like to say that sentence to a CEO or an IT manager in your government?

Fanboy comments like this are not enough to get Free Software into government offices.

Also a win for FreeBSD

[sremick]

I find it rather humorous that the poster of the article on Slashdot didn't dare mention the other software that was proven a winner by the Netcraft report. For those of you who haven't RTFA, 4th paragraph begins:

"Seven of the top nine sites run on FreeBSD." That's right, folks. NOT Linux. Don't get me wrong: I don't believe Linux sucks. But there's something to be said here by this data, and I don't feel Linux should get all the current press simply because Linux got all the past press. FreeBSD does amazing things, is used all over the place, has many technical merits not seen elsewhere, but Linux overshadows it because of inertia and the fact that Linux users yell louder. This is sad. Last I knew, Windows won out due to inertia as well, not technical reasons, and we condemn it for that. Must we be hypocritical a second time around?

I know this is Slashdot, but c'mon... would it kill you to put a positive article about FreeBSD on the front-page? ;)

Netcraft confirms it: FreeBSD is quite alive and kicking.

Backwards

[fm6]

I very much doubt it. People who dislike IIS probably aren't fans of Windows either. If they have a choice, they'll run Apache under Linux or Unix. If they don't have a choice, it's probably because the system is a personal workstation or a workgroup server. Which don't figure into the Netcraft numbers.

I think it's the other way around -- people choose Apache so they don't have to run Windows. It's probably not a coincidence that 2003 was also the year of the Windows Security Patch.

Re:Yes, but measuring webserver market share is ha

[dipipanone]

These surveys also do not count the millions of intranet-only sites that these servers serve

Are you sure you don't mean 'sites where administrator is too incompetent to turn off the default install of IIS'?

You know, all those sites that have plagued the internet with various worms and other security holes over the last few years?

and given the nature of the beast, I'm going to guess IIS is rather prevalent in that market

I don't disagree. I rather think IIS dominates at these sites.

Even higher praise for free software.

[twitter]

MS's recent campagin of Total Cost Of Ownership does not factor well into this.

Microsoft's TCO campaign is a last ditch effort to maintain market share. It's mostly a lie, but it's damaging to them even if true.

Assume they are telling the truth. I know that it's hard to keep a straight face reading that, but think of what it means. WHERE TECHNICAL MERIT IS THE DECIDING FACTOR, FREE SOFTWARE IS OVERWHELMINGLY PREFERED DESPITE HIGHER COST. Most companies ask themselves what a failed web site will cost them. The answer generally dwarfs the cost of the sofware and it's upkeep.

Of course, we all know that it costs nothing to aquire free web servers and less to keep them up than their non free counterparts. That's just the way good software works.

The same thing is true on the desktop. Most small businesses with a brain have a reseller to help them out with technical issues. Free software, when adopted there, will prove both cheaper and more reliable. Small businesses that dabble with HPs and their own M$ based IT are wasting time that would better be spent on their real business. The reseller may appear more expensive up front than trundeling down to CompUSA, but he's not. Resellers that move to free software are going to enjoy cost, feature and performance advantages that the 2003 server fanboys can only dream of. The same can be said of larger IT shops that can afford to do IT themselves, like ummm IBM [slashdot.org]

Microsoft's FUD campaign is running out of steam. They have tried all of this before and people are no longer listening.

It's actually even worse than that for Microsoft

[cabalamat2]

The figures you've quoted from my site [cabalamat.org] are accurate, but the situation for Microsoft is actually worse than that. When considering that Microsoft got 19% of new websites in 2003, it's worthwhile to consider that up to the end of 2002, Microsoft had a total of 24.74% of active sites.

This means that not only is Microsoft's share only 19%, Microsoft's market share is going down and Apache's is going up. Although Apache can run on MS Windows, it is nearly always run on Unix systems. The most popular Unix is Linux, which is busily replacing the proprietary Unices. So if Apache's share is going up (which it is) Linux's share is going up even faster:

Thus, for web servers at least, MS Windows is losing market share, and Linux is massively increasing its market share.

I don't have figures as to whether this is also the case for other types of server, but I strongly suspect it is.

Re:Yes, but measuring webserver market share is ha

[mce]

While I have no hard data to judge the intranet server market, I do know that at the place where I work Apache is used (on UNIX, no less), even though the majority of computer users over here use Windows. And I would expect a lot of places where computers are more important to the core business model than just being a comunication and/or database entry device to be doing the same thing. The reason for this is double:

1. Historical: Even though most people use Windows, those that actually know about computing using UNIX (for us, this used to be HP-UX, now it mostly is Linux). It are the latter ones who more than likely started the intranet effort long before management knew what a network was (over here, I myself was involved in our first intranet look-alike long before the word reached the trade-press).

2. Technical/Economical: If you use Apache for your external site (as we do), than it bloody well makes sense to use it internally as well, instead of wasting time and money maintaining two knowledge skills.

Intranet Reliability

[soloport]

They do not include the zillion corporate intranet servers...

Of all the intranets we install and service for small to large businesses, 100% of them run Apache. That's about 3-4 servers per month, and growing. We know 4 of the 5 competitors in our market, very well. For the vendors we know, all install Apache, exclusively.

Yes. Thanks to the "blaster outbreaks and the growing number of vpns", Apache is also rapidly growing inside the LAN market space.

Re:Also a win for FreeBSD

[mrroach]

> This is sad. Last I knew, Windows won out due to
> inertia as well, not technical reasons, and we
> condemn it for that. Must we be hypocritical a
> second time around?

Don't forget though, that linux "winning" != freebsd "losing"

Right now, linux/bsd are obviously Not Windows when it comes to the PHBs. If linux makes it to top-of-the-heap, freebsd is right there in line behind it. It may irk you that linux is not right behind freebsd instead, but don't pull down the friendly competition in the attack against the opposition. This town is big enough for the both of us ;-)

-Mark

Re:Yeah, but

[Anonymous Coward]

As Apache grows in use, so do the amount of security breaches.

How long has Apache been the market leader?
Seems strange that, as webservers go, IIS seems a lot more vulnerable.

Apache Jakarta: major in java world

[sundling]

The jakarta project (including ant which is now a top level apache project) has spawned a number of projects that are almost de facto in their area like struts, tomcat, cactus and many more.

Struts is an MVC framework that even includes tools to generate javascript validation code. This is a very common method to create a model 2 architecture J2EE site. Tomcat is the standard in open source servlet containers and often refered to as the reference implementation on a JSP and Servlet spec. Cactus is for unit testing J2EE components and is starting to become more popular.

If you intend to program java, then you should visit the Jakarta [apache.org] site.

As for who these people are, there are usually some pages on a project to mention that sort of thing. I'm most familiar with struts and their page for that sort of information is the volunteers [apache.org] page. Ant is already the defacto java build tool. Originally designed as a replacement for make, it's abilities can be extended using java classes.

Jakarta and Apache projects will continue to be a source of innovation, especially within the java world.

Paul Sundling

Re:It's tricky, alright

[Anonymous Coward]

That is not what he said. He said servers which have a larger install base will have slower growth in terms of a percentage of total install base. The comparison is meaningless. Apache has the largest install base of them all (bigger) and grew more (10 million), but only grew 40% of its total base. So its therefore growing slower than something that grew 1000 installations? No. I don't think so.

Re:Yes, but measuring webserver market share is ha

[abulafia]

But I distrust software monocultures, and I fear Apache's heading that way.

That's one of the nice things about Apache. Running Apache doesn't mean running the same Apache that someone else does. mod_perl, Jakarta, mod_php, mod_whatever are all competing with each other. Apache is essentially a platform, not just a server.

TCO is badly mis-understood.

[khasim]

TCO does NOT include migration costs (initial training and porting apps). (These are important factors and need to be addressed in the Return On Investment (ROI) calculations.)

TCO is NOT applicable between companies UNLESS they are practically identical (same number of techs with the same training managing the same number of servers with the same OS's running the same apps (not similar apps, the same apps) for the same number of users, connected in the same fashion (wireless, wired, VPN'd in, etc) using the same desktop OS, etc).

Usually, TCO will come down to human maintenance (and floor space, cooling, etc) and licensing costs.

Neither Migration Costs nor TCO take into account money lost when the server is DOWN!

Microsoft usually does the following:
#1. Incorrectly includes training for other products as TCO instead of Migration.

#2. With #1., they do NOT include training on Microsoft products (assumes people already know it).

#3. Ends the "period" prior to the NEXT round of license expenses.

The Migration Costs (plus) the annual TCO (minus) downtime savings = $$$ You have to get from ROI.

TCO is MEANINGLESS when used by itself.
-and-
TCO is usually calculated incorrectly anyway.

The REAL issue with Open Source is the MIGRATION COST because so many people have apps that they depend upon that must be ported.

Which is why Microsoft does tries to confuse the issue with bogus TCO claims.

If you focus on the MIGRATION COSTS, you can handle them in smaller chunks over a longer period of time. Bit by bit, move your systems over to Open Source based servers and services.

Re:Sends binary files as text/plain MIME type

[moof1138]

If there is a defined standard for handling unkown MIME, I am not aware of it. So what would you say is 'correct' behavior for a file with no registered MIME type?

RAR and Windows Media do not appear to actually have registered types AFAICT. I admit that sending unknown things as text/plain is a pain for users, but I think the solution is for all common file types to get registered as some MIME type, not to bitch at the webserver. Please feel free to correct me if you can find them here:
http://www.iana.org/assignments/media-types /

IIS sends unknown stuff as application/octet-stream, but that seems screwy to me - why treat an unknown file as a binary appication? If has an extension but it is not '.exe' chances are the file is not 'application/octet-stream'.

BTW - changes to Mozilla to check the file content for unprintable characters hit the trunk recently so Mozilla and Gecko based browsres will handle text/plain files that are not compliant to the text/plain type, and download them instead of displaying them.

is this really good?

[penguin7of9]

Yes, it's nice that Apache is open source, and it would be a disaster if the situation were reversed wrt. IIS.

But what I'd really like to see is a lot more diversity in web servers. Apache is a reliable, robust, efficient server, but it is only one, very specific way of serving web data and it has tons of quirks as well (starting with its configuration files).

Having Apache open makes it easier to innovate based on it. But I think it would be even better if more people did something altogether different rather than just plugging into Apache.

Um, sure, it's free

[Knights who say 'INT]

If I was to set up a web server, I'd use a Linux+Apache config too. But that's because I ain't got no money, not because I particularly trust Apache.

Re:Yes, but measuring webserver market share is ha

[JamieF]

Wow, Apache lets you add functionality through plug-ins that use a standard API? That's amazing! Just about every other web server has that too, but they don't make you run a bunch of command line config crap and recompile like Apache does, so it's not as k3wl. Recompiling is fun and definitely better than using some stupid installer that gets the configuration right the first time.

What other features can we gush about? Oh my god, it serves HTTP too? That's awesome! Can it talk to the filesystem and actually keep a log of the HTTP accesses, though? That would be really amazing.