Skip to content

[Bioc-devel] Bioconductor-mirror permanently removed August 25th

6 messages · Jim Hester, Turaga, Nitesh, Robert M. Flight +2 more

Original
Jim Hester
# 
This mirror could be relatively simply supported by setting up an
additional remote in the server side repositories and adding a post-receive
hook to each repository that simply pushes to the GitHub mirror remote
after commits are received.

https://git-scm.com/book/gr/v2/Customizing-Git-Git-Hooks
https://help.github.com/articles/about-github-mirrors/

Having an easily browsable source repository online is a pretty valuable
thing, it is a shame to lose this.

Jim

On Thu, Aug 17, 2017 at 3:53 PM, Turaga, Nitesh <
Nitesh.Turaga at roswellpark.org> wrote:

            

      
      
      
    

        

      
      
    

        

  
  


  


    

    

      
      

        


  

        

      Turaga, Nitesh
    

    

      
      #
    

  


  

      
Hi Jim

We are more worried about the confusion it creates among the users.

Users have tried to push their commits to the bioconductor-mirror instead of the bioconductor git server. Another source of confusion is, some maintainers have forked their packages from the Bioconductor-mirror, and have started developing from that point on GitHub. This creates a unique issue of duplicate commits, when they merge updates from the bioc-git server. People also send in pull requests/issues to this mirror repositories where they are unmaintained, causing a new kind of problem to both users and maintainers.

We realize the importance of a browsable source repository, and are looking into gitweb on the bioc-git server directly. Having a GitHub based mirror also prevents us from hosting data-packages on Github. The idea of a browsable source should be to view all Bioconductor packages( software / data / workflows). My vote is to remove this, only because of the amount of confusion it causes on the regular. 

I?ll let others chime in as well, and would love to hear your thoughts about Gitweb(https://git-scm.com/docs/gitweb) or other such software.


Best,

Nitesh

            

      
      
      
    

        

      
      
    

        
This email message may contain legally privileged and/or confidential information.  If you are not the intended recipient(s), or the employee or agent responsible for the delivery of this message to the intended recipient(s), you are hereby notified that any disclosure, copying, distribution, or use of this email message is prohibited.  If you have received this message in error, please notify the sender immediately by e-mail and delete this email message from your computer. Thank you.

  
  


  


      

    

    

      
      

        


  

        

      Robert M. Flight
    

    

      
      #
    

  


  

      
I agree with Jim. It has been very useful to have the Bioconductor GitHub
mirror to examine code, and the CRAN mirror as well. A post-hook to push
there would be very useful.

-Robert

        
On Wed, Sep 6, 2017 at 11:07 AM Jim Hester <james.f.hester at gmail.com> wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Martin Morgan
    

    

      
      #
    

  


  

      
On 09/06/2017 11:28 AM, Robert M. Flight wrote:

            

      
      
      
    

        
The mirror causes tremendous confusion on several fronts.

The updates can be not robust for a number of reasons (likely less so 
now that we use git).

The 'definitive' source is git.bioconductor.org, not the github mirror, 
and of course developers often have their own github repository. We 
regularly receive queries from developers trying to push to the mirror, 
users trying to post issues on the mirror, etc.

We have explored a couple of alternative approaches. It seems likely 
that the rss feed / svn log at

   http://bioconductor.org/developers/svnlog/

can be easily re-enabled. We've also looked at gitweb for browsing, 
which looks like

   https://root.cern.ch/gitweb/

and is 'different enough' from github that users / developers wouldn't 
get confused (and it would allow source-level browsing of the actual 
repository).

Further discussion welcome.

Martin

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        
This email message may contain legally privileged and/or...{{dropped:2}}

  
  


  


      

    

    

      
      

        


  

        

      Shepherd, Lori
    

    

      
      #
    

  


  

      
Even before the bioconductor git server, when we still were using svn, users were confused at where to make updates and push changes.  Now even more so we have seen several packages try and push to the bioconductor-mirror or change the remotes to point the the mirror locations.  It has created a lot of confusion.




Lori Shepherd

Bioconductor Core Team

Roswell Park Cancer Institute

Department of Biostatistics & Bioinformatics

Elm & Carlton Streets

Buffalo, New York 14263

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Jim Hester
    

    

      
      #
    

  


  

      
I understand the fragmentation concerns, just felt the mirrors should have
been maintained at least until a reasonable alternative exists.

The diverging history issues should be resolvable by resetting the state to
the canonical bioconductor repository and force pushing necessary changes
to their repository. Issues for the mirrors were disabled, and the
repositories had a pull request webhook setup to automatically close pull
requests with an informative message. You could potentially have similar
issues with the CRAN mirror (https://github.com/cran), but that seems to be
a valuable resource to the community.

Also FWIW the CERN root project repos are also on GitHub (
https://github.com/root-project), it is instructive to contrast the landing
page for the same repository on each
  - https://root.cern.ch/gitweb/?p=cling.gi
  - https://github.com/root-project/cling

I personally find repository interfaces that put commit messages up front
such as sourceforge and gitweb challenging to navigate. Generally if I am
looking at a repository it is because I want to see the source code, not
the latest commit messages. For self hosting solutions I would encourage
use of something like Gitea (https://gitea.io/en-US/), which has a richer
interface with easier access to source code / rendered READMEs etc.
(example repo https://try.gitea.io/klud/graphql-docker-api). I think
similarity concerns to GitHub are somewhat overblown and could be addressed
with some Bioconductor specific CSS if needed. There are performance
concerns with a richer interface, but I believe Gitea has much lower
resource requirements than other similar alternatives like GitLab and at
least their demo server seems to have no trouble with ~1300 repositories.

I feel the real value in having Bioconductor mirrors on GitHub is the
ability to easily search for code across all of the published R ecosystem
in one query. This is useful for deprecating functions, gauging usage
popularity etc. and becomes harder to do without a mirror.

Thank you to everyone for the replies and you are of course free to do what
you feel is in the best interest of the project!

Jim

On Wed, Sep 6, 2017 at 11:21 AM, Turaga, Nitesh <

        
Nitesh.Turaga at roswellpark.org> wrote: