jina-ai/jina-hubble-sdk

Parser for `jinahub+docker://name` doesn't work after introducing namespace

mapleeit opened this issue ยท 14 comments

We need to make the parser compatiple for both jinahub+docker://name/tag and jinahub+docker://username/name/tag

As we discussed before, current jinahub://executor1/tag1 format conflicts with
next name-spaced jinahub://username1/executor1 format

There is no way to check which format is being used on the client side.
How should we solve this @nomagick?
We must check all possibilities on the backend side, I'm afraid.

Will jinahub://username1@executor1 or jinahub://executor1@username be an option?

Try to learn something from https://en.wikipedia.org/wiki/Uniform_Resource_Identifier

Can we change the scheme? like jinacloud://username/executor?tag=latest

I don't like characters other than / ๐Ÿ˜„

Can we change the scheme? like jinacloud://username/executor?tag=latest

I like the idea of introducing a new schema (old one only works for old executors).

If we do it, how about just following docker pull jinaai/jina:master-py39-perf format?
Something like jinacloud://username1/executor1:tag1.

check all possibilities on the backend side

I don't think it's a good idea. There will be unsolvable cases like user HanXiao vs executor HanXiao.

I don't like characters other than /

Same to me.

jinacloud://username1/executor1/:tag1

I think this is ok.

But remember we have the syntax for private Executor: jinahub://exec_name:secret/tag

I remember when I first implemented this, I imaged jinahub://username1@executor1 as the way to introducing namespace.

I won't insist if you both think it's not good tho. But it's important to keep consistent with the old one in general.

We need either jinahub://username1@executor1 or jinacloud://whatever.
@delgermurun Could you make the decision? I'm ok with both. Or ask the team or Han for opinions.

Will jinahub://username1@executor1 or jinahub://executor1@username be an option?

@mapleeit What are the reasons for not using the second option? Are there any technical reasons?
If I had to choose one here, I would choose the second option because the executor belongs to the user rather than vice versa.

Then let's choose jinahub://executor1@username, make sense to me. I shouldn't cross out it. @delgermurun @nomagick

We will support both jinahub:// and jinaai:// prefixes, but only document jinaai:// one.

๐ŸŽ‰ This issue has been resolved in version 0.26.0 ๐ŸŽ‰

The release is available on:

Your semantic-release bot ๐Ÿ“ฆ๐Ÿš€