Re: [Bloat] Bufferbloat.net - Organizing, curating, and workflow

["Toke Høiland-Jørgensen" <toke@toke.dk>]

Dave Taht <dave.taht@gmail.com> writes:

> On Sun, Jun 12, 2016 at 10:49 AM, Rich Brown <richb.hanover@gmail.com> wrote:
>> To all the Bufferbloaters out there...
>>
>> Thanks again to Toke for all this good work. It's terrific to see the pages
>> resurrected, and to have the www.bufferbloat.net site living again.
>
> And it's fast - even half a planet away. I am still looking around at
> various cdn technologies....

I played around with using cloudfare through amazon for my blog at some
point; for static pages you can get it to pull from an S3 bucket.
Syncing to that is fairly straight forward.

I gave up on the effort at the time because I botched the SSL cert setup
(cloudfare only accepts up to 2048 bits, and this was pre-letsencrypt,
so getting a new one once issued was non-trivial), and because the
propagation delay for changes was too high.

We can certainly stick it in a CDN if you find something that works
well, but for now at least I'd call that a premature optimisation.

I did configure the web server to do most of the obvious speed
optimisations (for instance, everything will be statically compressed on
build, so the server basically just has to do sendfile() on the right
file on request).

https://developers.google.com/speed/pagespeed/insights/?url=https://www.bufferbloat.net/projects/


> Definately agree it would be good to track what people have been using
> as inbound links, on plain access also. People kept linking to the
> oddest parts of the site...

I can have the server generate awstats reports for the site
(http://www.awstats.org/) - these include the most common inbound
referrers and most common 404 pages. Not sure if they should be public,
though. Don't *think* they usually contain sensitive content (given that
the whole site is public, anyway). But I could be wrong; thoughts?

-Toke