I have more than 100 systems running ubuntu. I have a configuration file in one of these systems say 172...10. Can i get that configuration file from the other sytem say 172...100 using wget? I have installed wget? And i have tried the following wget http://172.*.*.10/name.conf but i am not able ...
As Ubuntu 10.10 seems to neither detect my graphics card (Intel 82852/855GM) automatically nor use corresponding Intel driver even after manually installing it, I am looking forward to manually configure X (shouldn't I?) Where an I find configuration files I need to edit?
@Alaukik Disqus is getting on my nerves. The numbers thrown around by Caitlyn Martin are fairly disparate in that they're a complete munge of silly sales estimates and feelings about how well Linux should be doing.
For everybody else, this is a continuation of a discussion under this thread about the real market share of Linux. I'm arguing for ~1% being roughly correct, @Alaukik higher
@Alaukik Extrapolation and pure guesses are wildly different. Her end number is so contorted by layering statistical flaws on top of each other, there's no value in it at all. At least with a single source, with no reasonable, foreseeable demographic bias (ie, the Wikimedia log parsings), you know roughly how many people are using what.
The only reason I got involved in this this time is I saw somebody mention 1% and somebody else accusing them of malicious (as inferred it) misinformation.
I'm not saying I like the 1% figure, just that everything else seems to spawn from fervid fairy tales about possible sales and not what people are actually using.
@Oli from wikipedia there are 2 billion internet users and from here (fedoraproject.org/wiki/Statistics#Total_repository_connections) there are 29 million fedora users . lets say 4 million users dual-boot (with any other gnu/linux disto) ubuntu (being a more end-user oriented distro and popular) must have most users than fedora lets say 26 million and then lets say 1 million users dual-boot so there are 50 million users of fedora and ubuntu so 25% of internet users use gnu/linux
it even states "Currently, there is no reliable way to determine the total number of Linux users, or even count the total number of users of any Linux distribution which does not have a mandatory per user registration process." → selective reading ftw! :þ
Oh I see it. 29 million unique IPs over the course of many years connecting to the repo.
I get about five different IPs per week. I know that's not the same for everybody but it's not uncommon for people to go more than a few a month. Repo-based stats are about as useful as sales stats because there are so many issues on both side of them. Reading that Fedora page would have told you that.
But yeah, even if the average rate was 1 IP per user per week, you need to divide that number by 260. You could then argue that some people NAT that IP through multiple machines. We've only been munging the numbers for a minute and they're already completely unusable.
Unfortunately, things have gotten busy at work just as its gotten busier here.... But we do what we can do with the Important Work of AskUbuntu Watching.
So, it seems like a reasonable strategy to advertise close votes in chat....
//Code sanitized to protect the foolish.
using System;
using System.Collections.Generic;
using System.Text;
using System.Reflection;
using System.Web.UI;
namespace Mobile.Web.Control
{
/// <summary>
/// Class used to work around Richard being a fucking idiot
/// </summary&g...
wget will only be able to download anything if there is an HTTP server running at the other end.
As you may be able to run commands via ssh in bulk, this might help:
# on the client machine:
cd /home/username/Pictures/
python -m SimpleHTTPServer
This starts a web-server on the machine it's ru...
what happened there :D
lesson learned, don't post the first answer, post the best one.
@fluteflute no, synonyms (and merging tags, but that doesn't happen unless there's something seriously wrong) only works on the whole set of question tagged with it, we can't "bulk-retag"
@fluteflute merging tags is the "no really, I mean it" version of synonyms, it's one of the very few things that aren't revertible , so no one ever does it :-)
I guess (and I mean guess, I haven't read up on it) the SE people think that tagging is supposed to be fuzzy, and that it's worse for the community to be nit-picky
@GeorgeEdison the code is full of docstrings explaining why it's so fast, but it best explain it, since I might have written some nonsense here and there :-)
I "lex" the tilte of the question, meaning I split it into words, taking care of punctuation and case (so, "Ubuntu One on Windows?" turns into "ubuntu one on windows ?" turns into [ubuntu, one, windows]),
then I put every one of those words in a hash-table, with the question ID as the value
then I do the same for every pair of words in the question, that's ['ubuntu one', 'one on', 'on windows']
I "lex" the tilte of the question, meaning I split it into words, taking care of punctuation and case (so, "Ubuntu One on Windows?" turns into "ubuntu one on windows ?" turns into [ubuntu, one, windows]),
@GeorgeEdison for every character that is not in the alphabet + space + numbers, I replace that character with itself enclosed in spaces (there might be a better way, like shlex, but I'm obsessed with O(1) performance :-)
for i in s:
if i.lower() not in "abcdefghijklmnopqrstuvwxyz ":
s = s.replace(i, " {} ".format(i))
result = tuple(i for i in ''.join(i for i in s.lower()
if i in "abcdefghijklmnopqrstuvwxyz ").split()
if len(i) > 1 or any(j in "01234567890" for j in i))
n = list(i for i in result if len(i) > 2 or
any(j in "0123456789" for j in i))
result += tuple(' '.join(n[i:i + 2]) for i in range(len(n) - 1))
return result
↑ this is the lexing algorithm
the penultimate line adds word pairs
I'll cut out a piece of the index so I can show you how it looks, just a sec
now, the set is stored similarly to a hash table, i.e. the memory location of the value set is (something like) the hash of the key, so, at the moment, looking things up in the index is a constant time operation
that is, both the index and the values of every key in the index have O(1) membership lookup
so, now we have the index and important_words, one is a hash-table (defaultdict) and one's a set. That's the whole initialisation process, now we get on to the meaty part
result = Counter()
for i in lex(query):
result.update(self.index[i])
there's more obviously ;-)
a Counter is just like a dict, only more concise to write
results.update(iterable) means
for every item in iterable, increment the counters value of the key 'item' by one
since looking up self.index[i] is a constant time operation, this is still O(1) (though for every word, let's say we're at O(k) where k is very small)
and this is the second part of search()
return {'questions': list({'question_id': i[0],
'title': self.title[i[0]]} for i in
sorted(result.most_common(maxsize),
key=lambda x: lcs_len(self.title[x[0]], query)))[::-1]}
Somehow I can't access the source code of Counter() on the python servers at the moment, but I'm pretty sure it's about O(n), and our n is onyl as large as maxize, which defaults to ten
this is actually the slowest part of the whole thing (asymptotically...)
when we have all of the maxsize most common items, we go about sorting them, using a score
now, the method is called lcs_len, but that's wrong (I didn't think to change it, I was doing longest common subsequence before)
@GeorgeEdison hey if you have it running, and if you have plenty of RAM left, try replacing the PORT constant's value with a list of ports and see what happens
@GeorgeEdison right, it's just for fun (they show up in different colours on the console), but I used it to see what the thing behaves like under heavy load
It'd be cool if you could make it work with your server (i.e. make the server accept the same type of URI), so we can compare them, maybe swap out parts
@Jacob say my server responds to an http request in 4 milliseconds on the local host (and that's on my laptop, so maybe 2 milliseconds on a proper server), how much longer will it take once it's on the public internet?
@ajmitch I couldn't test it on the Stack Overflow data, for lack of memory, but it's just as fast on the superuser database, which is about 78k questions I think (compared to our 10k)
@ajmitch yeah I'm only talking about speed, not memory. Unfortunately, I can't profile memory usage to save my life, but it's somewhere around linear, maybe n^(1/something)
@GeorgeEdison see that's what I didn't understand :-)
@ajmitch and in terms of numbers, indexing the SuperUser database takes 600 megs of memory; and then I can use however much I want for the cache, if there's any left :)
@GeorgeEdison As long as the server is up and reliable, I'm not going to insist on running this python stuff I hacked together in a day :-)
@GeorgeEdison this is the most important, if you get it right, all else should fall into place. it incorporates a few common problems with search algorithms: paste.ubuntu.com/592326
As Ubuntu 10.10 seems to neither detect my graphics card (Intel 82852/855GM) automatically nor use corresponding Intel driver even after manually installing it, I am looking forward to manually configure X (shouldn't I?) Where an I find configuration files I need to edit?
//Code sanitized to protect the foolish.
using System;
using System.Collections.Generic;
using System.Text;
using System.Reflection;
using System.Web.UI;
namespace Mobile.Web.Control
{
/// <summary>
/// Class used to work around Richard being a fucking idiot
/// </summary&g...
wget will only be able to download anything if there is an HTTP server running at the other end.
As you may be able to run commands via ssh in bulk, this might help:
# on the client machine:
cd /home/username/Pictures/
python -m SimpleHTTPServer
This starts a web-server on the machine it's ru...
