[tor-dev] First five Tor tech reports

[Karsten Loesing]

Hi Zack,

On 7/21/12 2:13 AM, Zack Weinberg wrote:
> On Wed, Jul 18, 2012 at 10:05 AM, Karsten Loesing
> <karsten at torproject.org> wrote:
>> On 7/18/12 5:12 PM, Zack Weinberg wrote:
>>> On Thu, Jul 12, 2012 at 6:12 AM, Karsten Loesing <karsten at torproject.org> wrote:
>>>> here are the first five metrics tech reports that I'd like to turn into
>>>> Tor tech reports (see #5405 for the idea behind this).
>>>
>>> I'd like to offer some typographical improvements -- it looks like
>>> these are being generated with mostly default LaTeX settings, which
>>> are not great for PDFs that people will mostly read online.
>>
>> Oh, that would be very useful.  Maybe we can come up with a template for
>> all Tor tech reports.
> 
> I have put together a suggested template as tortechrep.cls (it builds
> on the standard article.cls) and updated all five of your tech reports
> to use it.  I also made some very small adjustments to some of the
> content (notably the bridge descriptor listings).
> 
> You can get it from https://gitweb.torproject.org/user/zwol/tech-reports.git .

Looks really awesome!  Merged into the official tech-reports.git.  Thanks!

For other people reading this thread, here are the resulting PDFs:

https://people.torproject.org/~karsten/volatile/fivereports/

> The biggest bikeshed in here is probably the font selection.  I picked
> something that looks good to me and should be comprehensively
> supported in recent TeX Live, but the only thing I feel like
> _insisting_ on about the fonts is "not Computer Modern."

Fine by me.

>> Right.  There's always the trade-off between using colored graphs which
>> don't go well when printed and dashed/dotted lines which are at least
>> equally useful on screen and on paper.  I'm open to suggestions there.
> 
> I liked the varying shades of gray solid line that were in a different
> report.  (Some of those graphs also use dotted lines, but there's only
> one *kind* of dotted line and it's used for something that's not
> terribly zigzaggy, so it's probably OK.)

Oh, right, varying shades of gray would work.  Will try, unless you want
to tweak the graphs first.

>> (Note that the graph sources are in a different Git repository than the
>> LaTeX sources.)
> 
> Where do I find them?  I have a fair bit of experience with ggplot2
> and would like to experiment (may not be able to do so promptly,
> though).

You'll find the R/ggplot2 sources in the metrics-tasks.git repository:

https://gitweb.torproject.org/metrics-tasks.git

There's a task-xxxx/ directory for each report, with xxxx being the Trac
ticket number that lead to writing the report.  You'll be interested in
these three directories:

 - task-2911/ -- An Analysis of Tor Relay Stability

 - task-4030/ -- Case study: Learning whether a Tor bridge...

 - task-4255/ -- An Analysis of Tor Bridge Stability

You'll also need the .csv files which are not checked into the Git repo.
 I put them up here (7.8M):

https://people.torproject.org/~karsten/volatile/data-for-fivereports.tar.bz2

> If you change nothing else about the graphs, please redo the bitmap
> (PNG) graphs in bridge-blockings with vector graphics (PDF) instead.

Done.

> I'd also encourage you to experiment with tikzDevice; it's nice when
> the graphs are font-consistent with the main text, and using tikz
> gives you that for free.  Also R's PDF output is frankly pretty awful
> (most significantly, it doesn't embed any fonts, which makes some
> publishers very unhappy with you).  tikzDevice _can_ generate output
> that is painfully slow to compile through TeX, but I don't think these
> graphs will have that problem.

Sounds interesting.  Do you mind giving an example?

Thanks!
Karsten