Any Github Experts online ?

Hello NANOGers,

I have one customer that claims that 2 out of 17 downloads using the git
command on github's service are slow and poor on our network when compared
to others.

However, when not using the git command , but using a simple web page link
to a large zipped file from github, its always nice and fast. Using the
git command 8% of the time being slow is unacceptable. Github just doesnt
responds lethargically at best. BTW, have you seen how many hex digits a
github ticket number is ?

Of course Github says try a different ISP...Customer tries to tell me
comcast is better ! What ! I dont believe it. No help from Github NOC - we
have asked and asked... And we peer with Github and for some reason they
do not transmit the Prefixes of the IP range that the customer uses for
the git command. resolve IPv4 is not in the prefix list. So
the exit is transits.

I need more clues. Is it the resources the git command uses when checking
files for dates etc ?

Thank You
Bob Evans

If they are using 'git pull', or 'git push' for example, they may be
accessing the data via HTTPS or SSH.

Can your user do a 'git remote -v' and see if they are connecting via
HTTPS or SSH to assist you with troubleshooting?

Then see if it's something specific to one or the other and if it's
specific to github or all sites.


use GIT_CURL_VERBOSE=1 GIT_TRACE=1 git <command> will print out full
trace. you can also try other environment variables from

In my experiences, this is usually caused by MTU discovery issue.

Also, performing a git clone or git pull is significantly different than downloading a tarball. Someone correct me if I am wrong but Git essentially has to reconstruct the files from its internal data structures. If you clone a large repository with a lot of history it will take longer than a smaller repository with less history.