Suggest methods to pull data without cloning the complete EvalAI repo

docker/dev/fetching_docker_container.md

@@ -0,0 +1,42 @@
+### Alternatives to `git clone` to fetch data from EvalAI
+
+> Note: Some great points have been mentioned in previous PRs on the same topic.
+I found those solutions quite good, so I am including them along with the name
+of the person who originally came up with the solution.
+
+#### Maintain a separate codebase for the files used
+This can be a very useful addition. As the Django container
+is common to both EvalAI and EvalAI-ngx projects, it will be helpful to separate
+it from the EvalAI codebase. I would like to add here that this can also be
+useful in the EvalAI-CLI project where it can be used to setup the API server
+to perform integration tests with the server. (See [this PR](https://github.com/Cloud-CV/evalai-cli/pull/210))
+
+Suggested by: @yashdusing
+
+#### Use SVN to download required folders
+`svn` can be used to download folders from GitHub. A list of the required files
+and folders can be stored somewhere inside the ngx repo.
+
+The process involved is:
+1. In the URL to the folder, what is usually `tree/master` should be replaced
+with `trunk`. This will be the URL to be used in the `svn` command.
+2. The following syntax can be used to download the required folder:
+```
+svn checkout https://github.com/owner/repo_name/trunk/folder_name
+```
+
+Reference: https://stackoverflow.com/questions/7106012/download-a-single-folder-or-directory-from-a-github-repo
+
+#### Host a zip file containing the required data
+A zip file can be maintained with the required files. This should be updated by
+an automated script whenever an update is made in one of the related files.
+`git clone` will be run once on the server, the required files / folders should
+be moved and put into the downloadable zip file.
+This way, we can run `git clone` once and serve multiple times with it.
+
+#### Summary
+* While all three approaches are workable, I think the first one (recommended by Yash)
+is the most handy as it allows a lot of flexibility and opens up new possibilities.
+* According to the mentioned reply on StackOverflow, *GitHub takes around 30
+seconds to convert larger repositories.* Therefore, the feasibility of the second
+approach should be checked in practice.