Don't use the term "caching" in our docs

[arschles]

Since code/metadata isn't evicted, it's technically the wrong term

[robjloranger]

I think Brian mentioned too that 'registry' is the wrong term. The question then, how should we refer to Athens?

[robjloranger]

I think Brian mentioned too that 'registry' is the wrong term. The question then, how should we refer to Athens?

[robjloranger]

I guess it's really a repository in some ways. Not a version control repository, but a software repository.

An immutable module repository.

[robjloranger]

I guess it's really a repository in some ways. Not a version control repository, but a software repository.

An immutable module repository.

[arschles]

sorry @robjloranger - I worded this issue poorly. I was talking about documents like http://docs.gomods.io/intro/components/#proxy-athens - where it says cache public modules.

I think it's still totally fine to call it a Proxy, and this is mutually exclusive from the registry name - I was just referring to the word "cache" in the proxy documentation.

Could we call it a "datastore"?

[arschles]

sorry @robjloranger - I worded this issue poorly. I was talking about documents like http://docs.gomods.io/intro/components/#proxy-athens - where it says cache public modules.

I think it's still totally fine to call it a Proxy, and this is mutually exclusive from the registry name - I was just referring to the word "cache" in the proxy documentation.

Could we call it a "datastore"?

[michalpristas]

datastore seems fine 👍

[michalpristas]

datastore seems fine 👍

[robjloranger]

👍 datastore

[robjloranger]

👍 datastore

[glb]

I wish my first contribution to this project wasn't a dissenting opinion... From what I've read so far (and I recognize that you are the actual creators here) it seems perfectly valid to talk and think about Athens as a caching proxy. Where is it written that a cache is not a cache unless it evicts data?

"How long does Athens cache content for?" "Forever!" seems like a valid question / answer?

Caching is a natural word to use for what Athens does, and it seems a bit awkward to try to avoid it.

What am I missing?

[glb]

I wish my first contribution to this project wasn't a dissenting opinion... From what I've read so far (and I recognize that you are the actual creators here) it seems perfectly valid to talk and think about Athens as a caching proxy. Where is it written that a cache is not a cache unless it evicts data?

"How long does Athens cache content for?" "Forever!" seems like a valid question / answer?

Caching is a natural word to use for what Athens does, and it seems a bit awkward to try to avoid it.

What am I missing?

[arschles]

@glb thanks for commenting anyway, I'm glad you did, and welcome to Athens 😄

I think you're technically right, and I actually like "cache" a lot also. I also agree that something like an unbounded ttl-less cache is valid. My favorite part about the entire caching nomenclature is that we can use "cache miss" and "cache fill" - which perfectly describe what Athens does on the backend when a go get comes in.

The main problem we've seen is psychological. Some people have asked us questions (both on and offline) similar to "how can Athens be immutable if it evicts modules?" I think they're assuming that "cache" means ephemeral storage of some kind.

What are your thoughts?

[arschles]

@glb thanks for commenting anyway, I'm glad you did, and welcome to Athens 😄

I think you're technically right, and I actually like "cache" a lot also. I also agree that something like an unbounded ttl-less cache is valid. My favorite part about the entire caching nomenclature is that we can use "cache miss" and "cache fill" - which perfectly describe what Athens does on the backend when a go get comes in.

The main problem we've seen is psychological. Some people have asked us questions (both on and offline) similar to "how can Athens be immutable if it evicts modules?" I think they're assuming that "cache" means ephemeral storage of some kind.

What are your thoughts?

[glb]

Thanks @arschles ! The exact fit of the caching metaphor is what I like as well; I think calling Athens a datastore leads to more awkward questions about "how do you store stuff in it?".

Is this an opportunity to educate people that caches don't have to evict things? You could abuse the playful Q&A trope a bit with a section that says something like "Q: But don't caches evict things? A: Often, but not always! Athens is an awesome cache that assumes an infinite backing store and keeps things forever."

(You may detect a slight tinge of skepticism here, but I'm going to roll with the "builds must be reproducible forever" crowd and leave my "but projects get EOL'd all the time" hat elsewhere -- those projects can host their own Athens instance and it can get torn down with the rest of the project.)

[glb]

Thanks @arschles ! The exact fit of the caching metaphor is what I like as well; I think calling Athens a datastore leads to more awkward questions about "how do you store stuff in it?".

Is this an opportunity to educate people that caches don't have to evict things? You could abuse the playful Q&A trope a bit with a section that says something like "Q: But don't caches evict things? A: Often, but not always! Athens is an awesome cache that assumes an infinite backing store and keeps things forever."

(You may detect a slight tinge of skepticism here, but I'm going to roll with the "builds must be reproducible forever" crowd and leave my "but projects get EOL'd all the time" hat elsewhere -- those projects can host their own Athens instance and it can get torn down with the rest of the project.)