download statistics inflated by travis

[ygrek]

Is it possible to account for downloads from travis CS and substract from download stats?

[avsm]

Is it really worth the trouble? That number will also be inflated by web crawlers. If the downloads are that low, the number is well within the error margins...

On 27 Nov 2013, at 02:14, ygrek notifications@github.com wrote:

Is it possible to account for downloads from travis CS and substract from download stats?

—
Reply to this email directly or view it on GitHub.

[ygrek]

The crawlers will not usually download random archives and the majority of them can be diverted with robots.txt. Considering that every build of every reverse dependency on travis will download the package several times.. I guess it can substantially inflate the number..

[AltGr]

if we have a list of travis IPs, it should be quite straight-forward to filter them out during the log parsing.

[ygrek]

maybe they use some specific user-agent?

[avsm]

This is a total waste of time. Travis makes about 6 requests per-pull request and doesn't do so regularly at all. Is there any evidence of this inflation other than that?

On 29 Nov 2013, at 15:56, ygrek notifications@github.com wrote:

maybe they use some specific user-agent?

—
Reply to this email directly or view it on GitHub.

[ygrek]

I don't know really. I could try to investigate with the logs. Was just looking at extlib download numbers (for 1.6.0) and seen them grow rapidly in the first day after release, quite unexpected, trying to find the explanation.