npm, PyPi overtake CPAN

[๏̯͡๏ Guido Barosio]

Bumping up this thread, there were some problems on the npm side of the
world:

https://scalenpm.org/

http://blog.nodejs.org/2013/11/26/npm-post-mortem/




On Mon, May 27, 2013 at 5:45 AM, Anthony Lucas <anthonyjlucas at gmail.com>wrote:

> Fair point.
>
> Although, working closely with some fairly large CDNs in the past, I just
> *have* to point out that disaster / failure plans should never ever rely on
> them. Some companies seem to be confused on this.
>
> Seriously, CDNs are for speed, with the side effect (not feature) of
> sometimes masking outages.
>
> They don't solve them, they just make them harder to measure.
>  On Friday, May 24, 2013, Nicholas Clark wrote:
>
> 1) in an online environment, just a URL to a master site is a single point
> > of
> >    failure. Start to chain enough dependencies, and however reliable they
> > are
> >    at building and testing, it doesn't work if you can't download the
> code
> >    because one server is down
>
>
> How true is that these days though? PyPi isn't a single server that can go
> down. It's now running on a CDN[1], and even in the event of a failure of
> the origin, the CDN will keep serving last known good (much like the
> situation with a mirror and the master mirror failing.)
>
> Obviously it's not perfect - infrequently pulled traffic might not be on
> the CDN when a failure occurs (I suspect people on this list who work for
> fastey can provide us the exact details if they wanted) but the advantages
> of instant update of changes to all 'mirrors' are not to be sneezed at.
>
> I guess I'm saying its not as cut and dry as it used to be.
>
> Mark
>
> [1] as of about ten hours ago -
> http://mail.python.org/pipermail/distutils-sig/2013-May/020849.html
>



-- 
Guido Barosio
http://www.ted.com/profiles/1085580