Search squid archive

Re: TCP_MISS only

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 28/09/22 07:56, Andy Armstrong wrote:
Okay – but what happens if you are communicating with a non REST endpoint.

You are still communicating over HTTP. To interact with and benefit from HTTP agents like caches you need to comply to the HTTP semantics they use.

IMO, REST is just a useful tool to define (in abstract) an API's operation when considering what/how it needs to be implemented.


Consider a Web services endpoint for example where a request is only interacted with via POST but the operation for example may frequently be a read based function akin to a HTTP GET?

That is by definition a broken implementation of HTTP. The agent is using a *delivery* API (POST) for retrieval (GET).

If you can separate the delivery and fetch operations HTTP becomes much easier to use.


Is Squid just simply not going to help cache those requests?

Not *by default*, no.

POST implies changing some arbitrary resource *other* than the URL presented. Based on data and logic which may not be provided in the request message URL+headers.

To use POST with caching both the client *and* the server have to explicitly tell the HTTP cache agent(s) what to do on every single HTTP message.

- The client has to tell the cache whether a stored response is able to be produced as reply, what object-ID it is trying to retrieve, what object-ID's it already knows about (if any), and how old the stored object is allowed to be.

- The server has to tell the cache whether the response can be stored, what to use for a unique-ID of the reply object, how old it already is, how long it can be stored for, how and when to update it when it becomes stale.

The Squid refresh_pattern can provide defaults for the storage times when they are omitted. But all the ID related things and whether to use cache at all can only come from the client/server.


As you can see by limiting yourself to POST-only you have imposed a huge amount of complexity. Using GET instead for fetches makes all the above *optional* where now it is mandatory.


It is only helpful for more strict alignment to REST principles?


You lost me here. Squid implements HTTP.

REST is a very abstract simplification of basic HTTP/1.0 semantics. So the closer ones code aligns to REST the *easier* it is to implement HTTP properly. But HTTP/1.1+ are vastly more than REST.

HTH
Amos
_______________________________________________
squid-users mailing list
squid-users@xxxxxxxxxxxxxxxxxxxxx
http://lists.squid-cache.org/listinfo/squid-users




[Index of Archives]     [Linux Audio Users]     [Samba]     [Big List of Linux Books]     [Linux USB]     [Yosemite News]

  Powered by Linux