Apologies for cross-posting.
Hi all,
This is a quick headsup that I will be doing a short presentation at the Strata Online Webconference Data that Matters that is titled "Liberating data @ Wikipedia".
For more info: http://strataconf.com/strata-may2012?intcmp=il-strata-stny12-widget-link-to…
Best,
Diederik
Thanks Sumana,
I'd compiled all of my GLAM analytics notes and thoughts into a Google Doc
(noting that I'm not a tech-person, so it may not make much sense - but I
just happen to speak to museum technologists a good bit and had insights to
share). This was sent to the analytics team, but I never heard anything
back.
https://docs.google.com/document/d/1kptd2EUgZ5gSuh9PFP7SUl4wyL5JJGKp2eYxrQo…
I'll now add it to this page, maybe it will get more visibility.
On Fri, May 11, 2012 at 1:25 PM, Sumana Harihareswara <sumanah(a)wikimedia.org
> wrote:
> I know a lot of you want statistics, metrics, and analytics:
>
> https://lists.wikimedia.org/mailman/listinfo/analytics
>
> Diederik van Liere's inaugural post says it all:
>
> > Welcome to the the inaugural Analytics Mailing list email.
> >
> > Here all your analytics wishes comes true,
> >
> >
> > so proposals, ideas, crazy ideas, crazy crazy ideas are welcome here!
> > as long as we can count something it is welcome.
>
> So please feel free to join that, and to put
> https://www.mediawiki.org/wiki/Analytics on your watchlist! Also check
> out https://www.mediawiki.org/wiki/Analytics/Pageviews/GLAM . "The
> purpose of this page is to gather all the different documents,
> discussions and ideas related to GLAM Analytics and to define a list
> with the MOST important metrics. Please feel free to add any missing
> data, this is document is almost certain to be incomplete."
>
> --
> Sumana Harihareswara
> Volunteer Development Coordinator
> Wikimedia Foundation
>
> _______________________________________________
> glam mailing list
> glam(a)lists.wikimedia.org
> https://lists.wikimedia.org/mailman/listinfo/glam
>
--
Lori Phillips
Web Content Specialist | Wikipedian in Residence
The Children's Museum of Indianapolis
US Cultural Partnerships Coordinator
Wikimedia Foundation
703.489.6036 | http://loribyrdphillips.com/
Hey guys,
I'm trying to use the sha1 hashes of Wiki content for the first time (woot!
Props to D et al. for seeing it through) but I'm having some trouble
actually getting them out of the API/databases. It looks like the
checksums only go back to April 19th. Is this true of all pages? Is there
any plan to propagate the metric backwards?
-Aaron
(Moving this to analytics list, cause uhhh, why not?)
Ok some more info!
I was testing some of my changes to udp-filter on my test labs instance, and hey! whadyaknow! The test failed. One of the failures was due to the content type field not matching correctly. It looked like this:
text/html; charset=UTF-8
And actually, my test didn't catch the charset portion of this, my regexp is matching via whitespace. It was the semi-colon in the text/html string that caused the test to fail.
I started sleuthing, and determined that this only shows up in my logs if I am hitting mediawiki pages. I grepped mediawiki core, and found that there are tons of places that are manually setting the Content-Type header with the charset.
So I'm pretty sure the setting of the value of this header is coming directly from Mediawiki. But, that isn't really our issue. Headers are allowed to have spaces in the values.
Here's some example log output with spaces in headers.
https://gist.github.com/2648312
Summary:
Nginx:
- Does not escape spaces in Content-Type
- Does not escape spaces in Accept-Language
- Does not escape spaces in any header (afaict*).
Varnish:
- Does not escape space in Content-Type
+ Escapes spaces in Accept-Langage
+ Escapes spaces in any header (afaict).
Squid:
~ Removes charset from Content-Type header (?)
+ Escapes spaces in Accept-Language.
+ Escapes spaces in any header (afaict).
The remaining question we need to answer: Do we want to patch source code for these guys in order to fix this problem? I'd rather not if we don't have to. Do we have to? Uhhh, dunno!
-Ao
*As Far As I Can Tell
On May 9, 2012, at 11:42 AM, Erik Zachte wrote:
> For scripts it makes no difference: split("\t") is fine
>
> For browsing lines in an editor: I can disable tab expansion (or pipe to sed first)
>
> For human readability of pasted log lines in e.g. mail, it's no big deal (just slightly more ambiguous than a printable char)
>
> So, I'm fine which whatever works best, inter-process wise.
>
> Downward compatibility is another issue, but subscribers to our logs could get tabs replaced by blanks before data are forwarded.
> Anyway please don't change this overnight.
>
> --
>
> *KISS*
>
> Then again. Diederiks filter could easily and cheaply change all spaces to underscores, then all tabs to spaces, and no-one would even have to think twice, or make any changes.
>
> Erik
>
>
>
> From: Diederik van Liere [mailto:dvanliere@wikimedia.org]
> Sent: Wednesday, May 09, 2012 3:59 PM
> To: Andrew Otto
> Cc: Erik Zachte; Rob Lanphier
> Subject: Re: Using tab as delimiter instead of space in the log files
>
> Even I thought 12 months ago, "how hard can it be to use spaces only as delimiters" in a server log line. And 12 months later I now *know* it is extremely hard and those spaces will keep coming back, and we will keep discarding those log lines because they do not match the expected number of fields. Right now, these space issues are giving me headaches to come up with reasonable WikipediaMobile app usage for the Mobile team.
>
> But think a couple of months ahead when we want to stream our data into the cluster, do we really want a gazillion if else if statements to make sure that the data is in the right format before we store in the cluster? This will create a huge performance problem.
>
> By keep fixing spaces we will always behind the facts, because it takes a while before we discover the issue, and then we need to make changes to the config and get it approved or we fix it for once and for all.
>
> Because the user agent info is freeform I think it's safest to go with a tab control character but I am open for other suggestions.
>
> Best,
> D
>
> On Wed, May 9, 2012 at 9:40 AM, Andrew Otto <otto(a)wikimedia.org> wrote:
> > we have a lot of software scattered all over the place processing
> > this stuff (much of which is outside the purview of this group), and
> > we're probably going to break at least some of it with this change.
> If so ja, could be a problem. How widespread is this? We'd like to keep track of these uses going forward.
>
> >>> Spaces, tabs and any
> >>> non-URL character in a URL needs to be escaped by
> >>> Varnish/nginx/Squid/whatever
> >> the source can be located within hours and
> >> patched.
> > I really don't think we should start patching server software (like we did with nginx user agent string) to replace spaces.
> Diederik, I agree. If we can figure out how to always escape things properly in the log format config directives for each log source (nginx,squid,varnishncsa), then no problem. But patching and recompiling every time we find a new edge case like this would be really annoying and kinda dangerous. The more changes we make to upstream sources the more we have to remember to maintain and the more bug prone we make crucial software.
>
> Hmm, if we can figure out how to patch each of theses sources only once to make all whitespace characters in log fields escaped, then that might be ok. Mayyyybe, I dunno. What if we add a new source? Then we'd have to figure out how to patch it too.
>
> >> is there any printable
> >> char that could be used?
> Technically I'm sure we could find one, but I think we'd have more complaints about readability in emails and editors if we don't use whitespace than if we just used a tab.
>
> What if we used multiple spaces (2? 4? 5?) as a delimiter? I'm not sure all the log format config directives would allow us to, but this might be less prone to editor weirdness, and also might cause problems for fewer users. Then they can split(' '). Although, hm, I guess that isn't much better than split("\t"), eh?
>
> -A
>
>
>
> On May 9, 2012, at 9:21 AM, Diederik van Liere wrote:
>
> >
> > The tab would be inserted by varnish/squid/nginx when writing the log, we don't need to worry about URL transport or anything like that.
> > We keep running into new cases where spaces show up where we don't expect them, and I really don't think we should start patching server software (like we did with nginx user agent string) to replace spaces.
> > That locks us into specific versions of the software and the alternative is easy and robust.
> >
> > Erik: I don't think that editor settings are a convincing argument against the use of tabs.
> >
> > I am going to send this proposal to Wikitech mailing list.
> >
> >
> > D
> > On 2012-05-09, at 7:29 AM, Erik Zachte wrote:
> >
> >> I'm not strongly pro or con but here are some thoughts:
> >>
> >> I always get confused by tabs as they will be presented differently
> >> depending on editor settings (and different editors have different
> >> defaults).
> >> If someone pastes some log lines in an email, tabs might already have been
> >> expanded.
> >>
> >> Also Diederik's new filter will validate the line, and count fields.
> >> If it signals invalid lines, the source can be located within hours and
> >> patched.
> >> How many sources could there be? We have 3 streams?
> >> In theory I am in favor of validating at the earliest point.
> >>
> >> --
> >>
> >> I know this is tricky, but if we decide to change, is there any printable
> >> char that could be used?
> >>
> >> There are some characters that are out-of-bounds for urls, but some of those
> >> are unsafe for other reasons
> >> http://www.ietf.org/rfc/rfc1738.txt 2.2 section par. 'unsafe:'
> >>
> >> <quote>
> >> Other characters are unsafe because gateways and other transport agents are
> >> known to sometimes modify such characters.
> >> These characters are "{", "}", "|", "\", "^", "~", "[", "]", and "`".
> >> <unquote>
> >>
> >> One that comes to mind, but may have other quirks is 127/7F/
> >>
> >> Another 2 cents.
> >> Erik
> >>
> >>
> >> -----Original Message-----
> >> From: Diederik van Liere [mailto:dvanliere@wikimedia.org]
> >> Sent: Wednesday, May 09, 2012 3:33 AM
> >> To: Rob Lanphier
> >> Cc: Andrew Otto; Erik Zachte
> >> Subject: Re: Using tab as delimiter instead of space in the log files
> >>
> >> a) Not sure why this wouldn't solve the problem, the problem is that we
> >> consistently are encountering spaces in fields while we are also using space
> >> as delimiter. By changing the delimiter we are no longer affected by spaces
> >> in the fields. Varnish does not need to escape tabs, those tabs are
> >> configured in the varnish/squid/nginx config files. Andrew, can you confirm
> >> this?
> >>
> >> b) Who else is consuming the log files besides the analytics team and Erik
> >> Zachte? And changing to a tab as delimiter is going to be a small change, so
> >> yes it needs to be communicated to the people who consume these log files
> >> but not fixing it or alternative writing all kinds of patches seems a worse
> >> solution.
> >>
> >>
> >> My 2 cents.
> >> D
> >>
> >> On 2012-05-08, at 9:25 PM, Rob Lanphier wrote:
> >>
> >>> On Tue, May 8, 2012 at 6:06 PM, Diederik van Liere
> >>> <dvanliere(a)wikimedia.org> wrote:
> >>>> Instead of continuing patching our servers to fix these spaces issues I
> >> strongly suggest that we move away from the space as delimiter and start
> >> using the tab (\t) character.
> >>>
> >>> Eeek! I'm pretty sure that this will:
> >>> a) not really solve the problem. If the Varnish isn't escaping
> >>> spaces, what makes us think it's escaping tabs? Spaces, tabs and any
> >>> non-URL character in a URL needs to be escaped by
> >>> Varnish/nginx/Squid/whatever
> >>> b) we have a lot of software scattered all over the place processing
> >>> this stuff (much of which is outside the purview of this group), and
> >>> we're probably going to break at least some of it with this change.
> >>>
> >>> Regardless, we'll need to have a much wider conversation about such a
> >>> change than this group.
> >>>
> >>> Rob
> >>
> >>
> >
>
(cc'ing analytics for information)
http://toolserver.org/~dartar/reg2/
• merged the new account registration dashboards and the old Y2Y comparison into a single report
• added cronjob to refresh the Y2Y stats hourly
Hi folks,
Dunno if there's a private thread on this already, but I figured it'd
be good to utilize our underutilized public list for this.
I see Ariel has done some work on locke, presumably to deal with some
disk full problem that always seems to crop up at the worst time on
the weekend (thank you Ariel for being on top of this). Quoting from
the server admin log[1]:
07:03 apergos: manually rotates udplogs on locke, copying
destined_for_storage3 off to hume:/archive/emergencyfromlocke/ (jeff,
this note's for you in particular)
Andrew, this is probably worth following up on when you get a chance.
Thanks
Rob
[1] Server admin log: http://wikitech.wikimedia.org/view/Server_admin_log
Hey all,
I started reorganizing the analytics section of the wiki a while back and forgot to post an update. I've since:
- Created project-specific pages for our current work, especially the Reportcard and the processing cluster, Kraken.
- Reportcard: https://www.mediawiki.org/wiki/Analytics/Reportcard
- I moved the obsolete "2.0" documentation off the front page and into its own subsection. I also changed the URLs so they do not mislead by suggesting there is an API or a database involved. https://www.mediawiki.org/wiki/Analytics/Reportcard/2.0
- Added pointers to the 2012-13 budget planning that was going on.
- Added locations for sharing research papers and bookmarks. Unfortunately, I have no idea how to upload PDFs to the wiki, so that didn't happen.
Questions, comments, and suggestions welcome!
Cheers,
--
David Schoonover
dsc(a)wikimedia.org