From Wikipedia, the free encyclopedia

Contents 1 Old discussions 2 Criteria for direct inclusion 2.1 Popularity criterion 2.1.1 Main discussion 2.1.2 Google hits 2.1.2.1 Number of hits to qualify 2.1.3 IRC activity 2.2 Particular cases 2.2.1 Ubuntu/Kubuntu 2.2.2 Arch Linux 3 Edit button

[edit] Old discussions

• /Archive01

[edit] Criteria for direct inclusion

There's no clear definition of the criteria for direct inclusion in the template yet, so I'm writing what I think these should be.--Chealer 05:22, 1 May 2006 (UTC)

• Has to be an actual distro... LFS and Kubuntu are not distros.--Chealer 05:22, 1 May 2006 (UTC)
• Be popular. After discussion between NicM and me, we agreed that the method and limit used would be to include only the distros with >=2 M Google Linux hits (see the section about this below). --Chealer 22:12, 10 July 2006 (UTC)

[edit] Popularity criterion

[edit] Main discussion

• This is the start of the discussion about the popularity criterion and the last comments about it. Due to my early attempt to "archive" the discussion, this will probably be unreadable. Be popular. I propose top 10, since that should already cover over 90% of the market. I don't think it's realistic to set a different number since the basic criteria (market share of each distro and costs of adding a distro to the list) are hard to evaluate and a flamewar would be unavoidable. There are 12 distros currently. IMHO, Damn Small and Xandros could be removed, but I'm not fighting distrowatch about that. I removed PCLinuxOS, which is currently 10th at distrowatch. Please remember that distrowatch largely emphasizes new/"user-friendly"/hype/LiveCD distros. For example, PCLinuxOS is obviously not a top 10 Linux distro. So use your experience. --Chealer 05:22, 1 May 2006 (UTC)

After more discussion, there is some agreement between NicM and me that for this criterion, the method and limit used could be to include only the distros with >=2 M Google Linux hits (see the section about this below). This is not really applied to the template yet. If you agree on something between 2 and 3 million, please mention the exact number and feel free to apply it to the template.--Chealer 17:10, 9 May 2006 (UTC)

I just applied the results of this discussion with 2 M Google hits as a lower limit.--Chealer 05:46, 13 June 2006 (UTC)

Yes, 2M Google Linux hits is fine by me, so long as nobody disagrees within the next few days. If anyone disagrees in future they should hopefully bring it up here for discussion rather than unilaterally modifying the template. NicM 17:32, 9 May 2006 (UTC).

On another note, reorganising the talk page in mid-discussion make the diff cluttered and hard to follow. I know you may prefer not to read discussions using the diffs, but please bear in mind that others (like me :-) may want to and moving blocks of text around during an active discussion makes it hard to pick out what new text to reply to. It is also sometimes easy to make mistakes and accidently cut or edit people's text (not that I am saying you have done this). NicM 17:32, 9 May 2006 (UTC).

I do follow discussion using the diffs, and admit I could have commited the edits better. I also clearly did it at the wrong moment, sorry. I'll try to think about "archiving" this very last part of the discussion in a few days if there's no more activity to get clean criteria. --Chealer 05:31, 13 June 2006 (UTC)

Rereading this discussion I relase I may have come off as a little short-tempered, but I always seem to keep being interrupted right when I am doing this stuff. In any case, I apologise if that is so. I think it is a good idea to have a consistent measure for this template and am happy you brought it up in the first place and that you suggested Google, which had not occurred to me before. Let's just leave it a couple of days to see if anyone else floats in, and then implement it. Sound good? NicM 18:15, 9 May 2006 (UTC).

Yes. No need to apologize, temper is not reliable and this discussion had good results, although it could obviously have been less painful. --Chealer 05:31, 13 June 2006 (UTC)

• The following item was an early conclusion I made about the discussion. Judging from NicM's replies, it must have been too early :) This criterion is followed by a long opposition from NicM which I split in the following section. To summarize, after he understands my proposition, he ends up questioning the top 10 limit I was proposing, suggesting instead top 15. I disagree with this limit, since it lets distros with about 1% of the Linux distros market share in the list, which are IMO anecdotal enough to be moved to the extensive list. NicM also believes that PCLinuxOS should be in the list, although 2 people (1 being me) disagreed with that. He seems to need a third confirmation to accept the removal. Please confirm the removal of PCLinuxOS, and give your opinion on the ideal limit.--Chealer 00:25, 9 May 2006 (UTC)

You are incorrect. I do not think PCLinuxOS should be included, I think you should not come along and remove it without discussion and based solely on the fact you think it should not be there. NicM 00:53, 9 May 2006 (UTC).

Great, I just removed PCLinuxOS. If anybody wants it back in, please explain why it should be. BTW, if thinking that it shouldn't be here wasn't an appropriate basis for removal, what were the appropriate basis? Thinking that it should be here? Heh.--Chealer 17:10, 9 May 2006 (UTC)

I already said I am happy to see it removed when there is a consensus, not simply when we want to. 6 hours wait is not enough time. Leave it a 2-3 days at minimum and then if the discussion goes no further, apply the 2M criteria to entire list. NicM 17:27, 9 May 2006 (UTC).

Done.--Chealer 05:31, 13 June 2006 (UTC)

I did not say I wanted a top 15, I do not want a "top" anything. I think forcing it to a fixed number of distros is a poor idea. A line should be drawn based on some quantifiable and verifiable measure (a concept you seem to have some difficultly with), after which the template may be adjusted. NicM 00:51, 9 May 2006 (UTC).

I'm sorry if I wrote you said you wanted a top 15. You mentioned that an appropriate limit for you would leave about 15, which is what I should have written. You have to understand, though, that using a limit of the kind "X M hits" is only possible if we trust a single source of information. Otherwise, we'd have to create a complex method of rating distros combining several sources giving "Wikipedia editors combined Linux distro market share points" and establish a limit on this kind of points. It's needless to mention that just seeing the huge debate that already happened between two editors, making all editors agree on such a method would be either impossible or require way too much effort for this template. I have no problem with the concept of quantifiable (see below my comment on verifiable) measures. I do have a problem with your problem with people's experience, though. It took someone's experience to code Google and make it a reliable source, and it took someone's experience to mention Google hits as an interesting quantifiable source, and to establish that relying solely on them could be acceptable for this template.--Chealer 17:10, 9 May 2006 (UTC)

It took someone's experience to do many things, that is not the point. The point is that what we should use should be verifiable and repeatable. Having this discussion once is fine, but it would be nice to avoid having it many times. NicM 17:27, 9 May 2006 (UTC).

Again, see below my comment on verifiable measures (referring to Google). Indeed, it would be nice to avoid repeating this, so I'm glad we agree on a measure which is at least repeatable.--Chealer 05:31, 13 June 2006 (UTC)

I think a cut off point above 1 million Google Linux hits leaves the template too small. Perhaps we need to consider some more complex combination that includes several search engines, or Distrowatch, to leave us with a compromise list of around 12? NicM 00:51, 9 May 2006 (UTC).

Hum. This rather contradicts what you're saying later: "Combining the two is probably more effort than this template deserves". I'm quoting you about distrowatch and search engines, but I don't expect that combining several search engines would be any easier than combining Google and distrowatch. I'd rather settle for Google. 12 distros is not a problem for me (and is actually the number of distros I left before you started rollbacking me).--Chealer 17:10, 9 May 2006 (UTC)

It does not contradict anything I have said. It is a suggestion. I think we should use solely Google, I personally feel combining several sources is too much effort, but others may disagree. NicM 17:27, 9 May 2006 (UTC).

Well, if it was a suggestion, then OK, but I'd rather not follow it. Actually, there are two things I don't understand in your message two levels above. First, 1 M Google Linux hits seems to leave > 12 anyway, but you realized that later. The second is why suggesting to combine sources rather than lowering the barrier even more if 1M would have been too strict. But anyway, since we realized that 2 M is fine, it seems safe to skip clarifying this. --Chealer 05:31, 13 June 2006 (UTC)

Or even not become complex. A line at 2 million Google Linux hits would leave us with 12 entries. NicM 00:53, 9 May 2006 (UTC).

Right...I did mention that although I preferred 3M, 2M was acceptable for me. I'm glad to see we can now agree on 2M.--Chealer 17:10, 9 May 2006 (UTC)

2M is fine by me too. Let's leave it for a week to see if anyone else disagrees and then adjust the template. NicM 17:27, 9 May 2006 (UTC).

Done.--Chealer 05:31, 13 June 2006 (UTC)

Well, 11 at first, probably, but a few more may come or go. NicM 00:55, 9 May 2006 (UTC).

• Er, on the one hand you ask for a clear definition of the criteria, and explain it should be top 10 and based on popularity (top 10 where? presumably at Distrowatch, it is the standard measure of popularity), but then you remove distributions based only on your experience. Which is it to be? I think PCLinuxOS should be included. NicM 07:31, 1 May 2006 (UTC).

It seems you didn't understand the (pseudo-)method I suggested. I propose being a top 10 Linux distro as a criteria. Not top 10 at distrowatch or wherever (unless I'm unaware of some serious popularity evaluation). Just the (hard to guess) list of the 10 most popular and influential Linux distros. Using "experience" is the "method" I propose to evaluate if a distro is part of the top 10 or not. Of course, it's more a pseudo-method, or even not a method...because I don't know of a good method. It's possible to consult resources such as distrowatch, Google hits or netcraft, but we shouldn't base the top 10 on a single one. I hope this clarifies enough, otherwise feel free to ask for more clarifications.--Chealer 01:30, 3 May 2006 (UTC)

So, I'm removing PCLinuxOS again, and using it as an example of how to add to your experience: I suppose you readded it because it's 10th at distrowatch. Why do I disagree that it's part of the actual top 10? First, look at distrowatch's entry for PCLinuxOS: this shows that PCLinuxOS is a "user-friendly" distro, as confirmed by the official website's "PClinuxOS delivers Desktop Ready Software". As I said, "user-friendly"/LiveCD distros tend to be largely overrepresented at distrowatch. Then I check it's IRC channel. There were 32 people other than me on my visit, which is barely over 1/30 of #gentoo's 965 (Gentoo being distrowatch's #9). This is a bit unfair to PCLinuxOS considering #gentoo is Freenode's largest channel, but the proportion should be convincing anyway. Then I check Google hits: not quite 2 M, while Gentoo has over 30 M. Less than 1/15. Now, is it worth looking at netcraft for popularity as server? No, PCLinuxOS is a desktop distro and must be rare on servers. Finally, I look at my experience to verify that all these checks didn't lie to me. Do I know many people using PCLinuxOS? No, I'm not aware of anybody using it, which would be suspicious if PCLinuxOS was actually a top 10 Linux distro.--Chealer 01:30, 3 May 2006 (UTC)

This "pseudo-method" is far too vague. and even if it is used it must be the product of some sort of consensus. You cannot just decide it and apply it, particularly when I have already said that I disagree with your choices, and nowhere near sufficient time has passed to see if anyone else has an opinion or complete the discussion. NicM 07:44, 3 May 2006 (UTC).

Sure, but who's opposing to the consensus yet? You initially misunderstood what I was proposing, but I don't think you disagreed with it yet, you only said you disagreed with my distro choices. Well, the sufficient time is hard to determine. Of course, there hasn't been a long time, but for something as obvious as PCLinuxOS being or not in the top 10 Linux distros, I don't think any time is needed.--Chealer 07:59, 4 May 2006 (UTC)

I'm letting PCLinuxOS for now, since it's part of the top 10 according to one of our sources, but please remove it. It was initially added at the article creation, when the article listed no less than 18 distros. Someone later removed it, and you readded it. I then removed it, after which you rereadded it, I reremoved it, and you rerereadded it. So, you seem to be the only one thinking PCLinuxOS is a top 10 Linux distro.--Chealer 07:59, 4 May 2006 (UTC)

I think we should take the top N on Distrowatch, which is a simple, confirmed, unarguable list, whatever its flaws. If you can find other sourced lists as well, we can take some amalgam. I suspect IRC channels are poor evidence, popularity of a distros and "how many people you know" are purely anecdotal and may be explained by other factors (eg, SuSE is more widely used in Europe than the US, a person in the US has a higher chance of not knowing anyone who uses it than someone in Europe; Gentoo favours the technically competent and tinkerers, its IRC channel being large is not a surprise and its freenode channel is the primary IRC channel for the distro. FreeBSD's channel on freenode has 294 people to OpenBSD's 172, yet no other measures indicate OpenBSD is 60% as popular as FreeBSD: Distrowatch says 23%, surveys say around a third. However, neither have an official channel on freenode. On other networks the story is different again, OpenBSD's efnet channel has arount 66% of the population of FreeBSD's.). NicM 07:45, 3 May 2006 (UTC).

IRC channels activity is actually fairly representative of the actual popularity. See the list below. It contains only one distro not in the current list (Arch Linux), and 2 which should perhaps be removed (CentOS and Slackware). Depending on who you're talking about, "how many people you know" can be anecdotal, but of course we don't expect clueless people to edit this article. The geographical repartition of users doesn't matter as long as you have a fair contact network on the Internet.--Chealer 07:59, 4 May 2006 (UTC)

I already demonstrated an instance in which it does not appear to be representative at all and your rankings below agree (Gentoo top IRC but low on Google? SuSE/Red Hat both nowhere IRC but near top of Google?). NicM 14:48, 4 May 2006 (UTC).

I fail to see how you've "demonstrated an instance in which it does not appear to be representative at all". According to Google hits, FreeBSD would be about 25 times more popular than OpenBSD, which is about 8 times more than you indicate it should be. Meanwhile, the IRC channels hint that FreeBSD would be about 1.7 times more popular than OpenBSD, which is about 2 times less than you indicate it should be. So, do we have to conclude that Google hits are worthless because they're 4 times more out of order than IRC activity in your example? No, it's just an example. We have to look at the results in general, and as I said, IRC channels give a top 10 at least as accurate as distrowatch's. --Chealer 04:30, 5 May 2006 (UTC)

I don't know where you are getting your Google hits but my Google says OpenBSD is about 1/4 to 1/3 of FreeBSD[1][2], which is a lot closer to what surveys and Distrowatch say than the 60% on Freenode IRC channels. NicM 01:04, 9 May 2006 (UTC).

Huuum. Good point. I can't figure out where I was getting that neither. This was a few days ago, but I'm pretty sure that was from Google BSD, which now shows 6,690,000/2,660,000, which is about 2.5, essentially what you suggested could be expected. I'm sorry about this. There's currently no way for me to make sure I didn't do an incorrect search or analysis. I'll try citing sources when it's easy like in this case in the future. On the other hand, you haven't defined what would be a representative proportion. The IRC stats for this case is clearly less accurate than Google hits, but does that allow to generalize that a bias of proportion above 6/5 for one case make a source worthless? Hardly. I think the discussion about IRC is getting boring, and we'd rather kill it here. Perhaps we can remember that the variability of Google hits may be problematic in the future (or that I'm bad for calculations).--Chealer 17:10, 9 May 2006 (UTC)

Everything is unreliable, unless by some miracle every distro started suddenly collecting usage statistics in the same way. I just think IRC is more unreliable. NicM 17:47, 9 May 2006 (UTC).

Right, we established fairly well than IRC is more unreliable than Google. I propose again to kill this discussion here. I'm sorry for bringing it in the first place with my obviously erroneous calculation about OpenBSD vs FreeBSD Google hits.--Chealer 05:31, 13 June 2006 (UTC)

If there are to be rules, they must be based on measures that anyone, clueless or not, can look at and agree that a distro should or should not be included. Ad-hoc measures like "how many people you know" or "size of IRC channel on one network" is not good enough. So far we have Google hits and Distrowatch as acceptable measurements. NicM 14:48, 4 May 2006 (UTC).

Unfortunately, we have to come up with the most important Linux distros if we are to keep this template. We can't list all 100+ Linux distros with it. We can either try to determine the most important Linux distros, or scrap the template right now and remove it from the pages that include it. I'm not excluding the possibility to scrap the template, but this discussion is in the perspective of keeping it. Now, we have to come up with this list. Basing exclusively on IRC channels would be a bad idea, as you say. Basing exclusively on any editor's personal experience would need a skilled editor. It's clear that distrowatch's ranking is inexact enough. Now, it may be that Google hits will be a reliable ranking, and we can adopt that without relying on anyone's experience...let's try that. Just let me clarify though, that although I probably could have been more clear about that, when I first talked about experience, I didn't mean that experience==number of people you know. Your experience is your knowledge of the Linux world. I mentioned that I knew no one using PCLinuxOS because that made it obvious that it wasn't a top 10 Linux distro, but that's only useful to fix gross errors. "Experience" in this context includes the Linux users you know IRL, on the Internet (mostly), the number of servers running the distro you saw, the number of mentions in the media, in IRC conversations, in real life conversations, etc.--Chealer 04:30, 5 May 2006 (UTC)

I think you are failing to understand me. It doesn't matter if you are Linus Torvalds himself, "experience" is not an good way to pick and choose Linux distributions for this list. No matter his knowledge, using one person's experience is an ad hoc, approximate thing that many people will disagree on. Hell, we have pretty much been using experience so far and you have come along and disagreed with it. NicM 07:24, 5 May 2006 (UTC).

I don't see why you call someone's experience "ad hoc". What less ad hoc was suggested? Google hits? Is that less ad hoc than someone's experience? Anyway, arguing against something saying that it is ad hoc won't help you. "ad hoc" is about the vaguest reproach you can do. Obviously, using someone's experience is approximate. Is that opposed to an accurate measure which would be possible? Of course, many people will disagree if we don't get an appropriate reference to choose the distros...but which appropriate references do you suggest then? I already explained that if we have no acceptable way to produce the list, we'll have to eliminate the template. So far, we don't know if experience or anything else was used, but that doesn't matter, as we didn't have criteria so far. I haven't disagreed with anything, I simply proposed criteria and partially applied them. The only person that could be said to disagree yet is you by your readdition of PCLinuxOS, but we have quite clearly proved that you were wrong to do that.--Chealer 00:25, 9 May 2006 (UTC)

No. Just because the criteria are undefined does not mean you can come and arbitratily enforce your own. NicM 01:15, 9 May 2006 (UTC).

Right. So, if you have problems considering that I was enforcing my own criteria, feel free to consider that I wasn't doing so, but simply editing the article the same way it was done before, without using any defined criteria.--Chealer 17:10, 9 May 2006 (UTC)

Look, its quite simple: you removed it, I disagreed and reverted so you. So, you took it to the talk page which was great, but you should then wait until it is clear that the discussion is over and a consensus is reached before making the change again. The point is to start a discussion instead of a revert war, not in parallel to it. Two people does not consensus make, it is simply good manners to wait for a few days and see if anyone else has an opinion, if they did, they probably would have said something by now, but it is only polite to give them a little time before implementing whatever we conclude. NicM 18:05, 9 May 2006 (UTC).

I admit that I wasn't "polite" in this. I considered PCLinuxOS too minor to be polite about that, but having it or not in the template is similarly minor, so I'm sorry if I should have waited for a consensus longer.--Chealer 05:31, 13 June 2006 (UTC)

You tried to remove it, I disagreed, with perfectly good reasons (it is high on Distrowatch), so I restored it. NicM 01:15, 9 May 2006 (UTC).

Even if being 10th in distrowatch's 6 months statistics was being high, it wouldn't be a perfectly good reason. I have already explained why distrowatch was a poor indication of popularity for new/LiveCD distros. If you want to argue this, please go back to the "thread" we discussed this in.--Chealer 17:10, 9 May 2006 (UTC)

Just because you think it is a poor reason does not mean that I have to think so too. Distrowatch is a common method of measuring Linux popularity and the most usually cited up to now, so removing it on that basis was perfectly fair. I happen to agree it is probably biased towards the views of hobbyist/home users rather than server users, but I haven't got any evidence other than it appears "obvious" either, or that that is necessarily a terrible thing for this template. Distrowatch is a measure of popularity, nobody has ever claimed it is definitive or that it is the only measure. Anyway, let's forget about Distrowatch, Google has always been a much better idea. NicM 17:47, 9 May 2006 (UTC).

I still disagree that readding PCLinuxOS based on distrowatch ranking was "perfectly fair" (at least, doing so after the discussion we had about DistroWatch), but this is past and I agree that it's better to move to something else.--Chealer 05:31, 13 June 2006 (UTC)

When a consensus has been demonstrated that it should not appear, or a selection process agreed to that it does not fit, it may be removed again. I have already explained which measures I find suitable several times: those that are simple and easily verifiable, that anyone can look at, repeat and demonstrate, regardless of their personal feelings or experience about what should appear. We have two of those so far: Google (or similar search engines) and Distrowatch. Both are an often-quoted measure of popularity. Combining the two is probably more effort than this template deserves, so I suggest picking a suitable point on one of them and including all distros above it. You may list the template for deletion at any point if you like, whether that effort will succeed I do not know. NicM 01:15, 9 May 2006 (UTC).

I agree to settle for the Google hits. I agree that combining sources would require more effort than this template deserves. This doesn't mean that experience is a bad criterion though. We did need experience to establish the pertinence of Google's ranking. As for calling Google hits "easily verifiable", this is not quite exact. Google hits do have a limited variability. But we have no way to verify if they match the actual market share.--Chealer 17:10, 9 May 2006 (UTC)

Maybe I'll explain this a bit more clearly: it must be something that when someone comes along in 6 months and says "I think Bob's Linux Distro should appear" we don't have to say "oh, in my experience it isn't popular enough," and get into a yes-it-is-no-it-isn't debate. We can say "does it get 2 million Google Linux hits? Yes? It's in. No? It's out." NicM 01:21, 9 May 2006 (UTC).

Right, that would be ideal if we have a reliable source that allows us to proceed this way. If you agree that Google has such a reliability though, you'd rather start applying the results than go on denying the value of the experience of other people.--Chealer 17:10, 9 May 2006 (UTC)

Google is probably not terribly reliable, but it is probably the best indicator we are going to find. An amalgam of sources may be more reliable but is probably overkill, Distrowatch does probably have weaknesses, perhaps we could dig up surveys or something, but is is really worth the effort? Google gives a simple, repeatable, verifiable number that is based on a large amount of data, not on a single person's feelings, experience or lack of experience. For this purpose there is little value in a single person's experience, in the same way as there is no value in predicting election results by a survey of one voter. NicM 17:54, 9 May 2006 (UTC).

See my comment above about Google numbers being verifiable. As for the comparison between a survey of one voter and elections, this is not a fair comparison. A survey of one voter in elections could compare to replacing all the distros in the template by the single distro some random person would use. Using our best editors' full experience is vastly better than this.--Chealer 05:31, 13 June 2006 (UTC)

You and I could both write a list of top 10 or 15 Linux distros from our experience. Would they be the same? Mostly perhaps, but probably not entirely. You even admitted yourself that you were very surprised by the popularity of CentOS. NicM 17:57, 9 May 2006 (UTC).

Indeed. One thing this highlights is that the lower we put the minimum market share, the harder it will be to determine if a specific distro meets it or not. If we would go for 0.5% share, it would be very hard to determine the distros meeting it from experience.--Chealer 05:31, 13 June 2006 (UTC)

I have also yet to see any evidence Distrowatch favours Live CDs and desktops as you say. NicM 07:45, 3 May 2006 (UTC).

This should be quite obvious. Just looking at the top 10 for the past 6 monthes: MEPIS, 5th, Damn Small, 6th, and KNOPPIX, 8th (and PCLinuxOS, if you wish, 10th). Meanwhile, Debian is in the middle of those, 7th. Heh...--Chealer 07:59, 4 May 2006 (UTC)

You are almost getting close to somewhere with Google hits, although they do have a a bias in the case of older distros and distros which have changed name over the years. NicM 07:45, 3 May 2006 (UTC).

The varying age of distros make it harder to determine a distro's current market share using Google hits, but this is hardly a bias. A distro's historical popularity matters in the decision to include it here or not.--Chealer 07:59, 4 May 2006 (UTC)

Why don't we select a number of Google.com hits on the distro's full, current name and a Distrowatch ranking that together make a distro eligible for appearance? I would suggest perhaps 2 or 3 million Google hits ("gentoo linux" has 5.5 mil, "suse linux" 19 million) and top 20 on Distrowatch? This is only a guess, however, I don't have time to Google all the current list to see which ones would pass in this case. NicM 07:44, 3 May 2006 (UTC).

Er, I was using google.co.uk for those numbers, but the principle is the same. NicM 07:45, 3 May 2006 (UTC).

Done, see "Google popularity". 3 M sounds fair.--Chealer 07:59, 4 May 2006 (UTC)

• You'll also note that in previous discussions people have permitted CentOS and PCLinuxOS to pass through, so I guess their experience is different from yours. I've restored them for now pending some more opinions on the matter. NicM 07:34, 1 May 2006 (UTC).

Right. On the other hand, LFS and Kubuntu were listed before I started watching this article, so I guess we shouldn't take the previous versions too seriously. As I said, I'm removing PCLinuxOS, but I'm letting CentOS, as it surprised me in Google hits, and particularly IRC channel crowding.--Chealer 01:30, 3 May 2006 (UTC)

[edit] Google hits

Number of Google hits in millions, rounded by million, of the Linux distributions from distrowatch's top 20. Taken from Google Linux at 20060913. Numbers can vary.

1. Debian (22)
2. Red Hat Enterprise (1.5, "RHEL":+0.8, "redhat":+17) Estimated ranking ("redhat"'s relevance hard to evaluate)
3. SUSE (15)
4. Fedora (12)
5. Ubuntu (9)
6. Mandriva (6, "Mandrake":+6) Estimated ranking
7. Slackware (8)
8. Gentoo (6)
9. Knoppix (4)
10. CentOS (2)
11. MEPIS (1.7)
12. Puppy (1.5)
13. Xandros (1.2)
14. Damn Small (>1)
15. Slax (>1)
16. Arch ("Arch Linux":<1, "arch":+5) Estimated ranking
17. PCLinuxOS (1)
18. Vector ("Vector Linux": <1, "vector": 3) Estimated ranking ("vector"'s relevance hard to evaluate)
19. Kanotix (<1)

[edit] Number of hits to qualify

The following discussion intended to determine where to set the limit for distros to include. There were subsequent discussions, and the results are summarized at the beginning of the article.--Chealer 05:33, 13 June 2006 (UTC)

Given these numbers I would go with anything at or above 1 million and forget about Distrowatch completely. The template needs 10-15 entries to be worth the bother. If the 1 mil number leads to far too many it can be increased. If there is doubt about whether an entry is above this threshold, I'd probably leave it out. NicM 14:48, 4 May 2006 (UTC).

Actually, with 9/10 entries, the template's content fits easily on 1 line instead of 2. 1 M is IMO too generous, leading to the inclusion of Xandros (currently included, but doubtful), Damn Small (currently included, but doubtful) and Slax (certainly nor worth being mentioned). As I said, I consider 3 M fair. 2 M is also possible. Taking 1 M would lead to the inclusion of distros which, according to Google hits, have about one hundredth of the Linux distros market share, which is becoming anecdotal. --Chealer 02:21, 5 May 2006 (UTC)

I think two lines is better than one. This template is not meant to be a hard and fast list of the top N Linux distributions, they do not get prizes for appearing or anything. It is meant to aid people reading about Linux distributions or navigating Wikipedia, I think we need to provide enough that it is actually useful. I would reckon 1 million would leave around 15 or so, which is a good number (not too many and not too few), even if we end up with a couple like Slax or Xandros that aren't really important. It would be nice to include Damn Small, it is a unique distro even if it is not widely used. NicM 07:24, 5 May 2006 (UTC).

If I understand you right, you want to forget about distrowatch completely and go with Google hits, which clearly indicate that PCLinuxOS is not a top 10 Linux distro (and apparently not a top 15 one neither). I consequently expect that you'll remove it, or explain why you'd still think it should be included. --Chealer 04:30, 5 May 2006 (UTC)

After we decide upon a system and have left sufficient time (a week or so) for othere to voice their opinion, if they have one, PCLinuxOS may be removed if it doesn't make the grade. NicM 07:24, 5 May 2006 (UTC).

[edit] IRC activity

Largest Linux distro channels on irc.freenode.net's at 20060504.

1. Gentoo (926)
2. Debian (743)
3. Ubuntu (692)
4. Slackware (##slackware, 249)
5. Fedora (235)
6. Arch Linux (165)
7. CentOS (154)
8. SUSE (150)
9. Mandriva (113)
10. Red Hat Enterprise (68)

[edit] Particular cases

[edit] Ubuntu/Kubuntu

Would anyone object to turning ubuntu into (k)ubuntu (that is [[kubuntu|(k)]][[ubuntu linux|ubuntu]] )? I think kubuntu doesn't register on the google hits/distrowatch/etc because it gets lumped together with Ubuntu. Rebuttal to the obvious questions about edubuntu and ubuntuX: Edubuntu is not and does not aim to be a widely used distro as it has a very specific purpose. UbuntuX is too new to say anything about. If you look on the ubuntuforums website they have an (equal) division between ubuntu and kubuntu and very little space given to edubuntu and ubuntuX. The bellman 12:30, 29 May 2006 (UTC)

Object. They are not lumped together in Distrowatch... What I would consider is to have Ubuntu be a "main" distro and Kubuntu et al in the More... category. But that's just IMHO, so bombs away. elpincha 12:41, 29 May 2006 (UTC)

[edit] Arch Linux

I removed Arch Linux, added by Samsara, according to the popularity criterion discussed above. However in this case the " <1 " number is an estimation I did due to the usage of the word "Arch" not being exclusive to Arch Linux. <1 is a quick estimation I did. It should be >=2 to be included. I don't think it's worth it, but if anyone thinks Arch Linux is >=2, a new estimation can be discussed.--Chealer 05:38, 13 June 2006 (UTC)

[edit] Edit button

I think the edit button should be removed. It's too tempting for newbies to put their favorite distro on the list. IMHO, if you're not WP-literate enough to edit the page you're on and see the name of this template and come here yourself, then you probably won't bother reading the inclusion discussions on this talk page. I'm removing it but willing to discuss. — RevRagnarok ^{Talk Contrib} 10:47, 29 August 2006 (UTC)

I think it is much more in the spirit of Wikipedia to have an edit button. A large comment at the top of template source pointing out the criteria and that the talk page should be read before adding entries would perhaps be a good idea. NicM 20:52, 31 August 2006 (UTC).

I agree editing is the spirit of WP, but how many templates do you know of that have an edit button? How many don't? — RevRagnarok ^{Talk Contrib} 22:48, 31 August 2006 (UTC)

Dunno. Some do, some don't. It doesn't really matter that much to me whether we have one or not, and it certainly looks neater without. NicM 14:25, 1 September 2006 (UTC).

It is usually only templates containing data that are ment to be updated in a regular manner that have an edit button, it is seldom that a template ment to be shown in main namespace to have an edit button, mostly only maintenance templates have those (for example {{todo}}) _→AzaToth 14:03, 27 September 2006 (UTC)