CacheTagHelper should not depend on IMemoryCache

[sebastienros]

IMemoryCache is a perfect match for single node scenarios in CacheTagHelper but it breaks if the website runs on more nodes, as each node would return different contents. The issue here is that there is no way to change it.

I think it would be better to either:

• Depend on IDitributedCache as the default implementation uses IMemoryCache which means it would work the same as today on a single instance, and not be broken on multiple instances.
• Or even better, define a custom interface for extensibility reasons so that each customer could decide how to store the cached tags: explicitly IMemoryCache, IDistributedCache or another implementation. Each site is different, and locking down the cache tag storage might be an issue. And having the to share the same implementation for controllers and tags might not make sense either.

A different approach would be to use a message bus implementation to invalidate IMemoryCache entries across nodes, but I don't think it's been planned either.

/cc @NTaylorMullen

[Eilon]

I agree with the thought in general, but I wonder if this is a case for maybe having different tag helpers? It could be argued that this tag helper is more of a <memoryCache> tag helper as opposed to just <cache> and what's described here is some other tag helper. Or not...

@DamianEdwards any further thoughts?

Parking in the Backlog for now.

[DamianEdwards]

I think the cache tag helper should have an attribute to let you select between im-memory and distributed cache.

[Eilon]

Unfortunately that breaks DI 😦 🐼

It would mean you can't take the cache interface in the ctor anymore: https://github.com/aspnet/Mvc/blob/dev/src/Microsoft.AspNet.Mvc.TagHelpers/CacheTagHelper.cs#L47

You'd have to reach into the service provider and resolve it.

That's the real reason I was suggesting multiple middlesware.

[sebastienros]

You can take a dependency on both.

[Eilon]

But it doesn't actually require both, so that's a bit dirty.

[sebastienros]

If Damian's advice is to have an attribute to target which storage for each tag (which is what I understood), then it actually requires both services. You can make them lazy if you want, still, not a broken design IMO.

[pranavkm]

Supporting IDistributedCache would require us to serialize the generated IHtmlContent to a byte[] which is entirely incompatible with how we cache it in IMemoryCache.
What's the use case for caching generated HTML content as opposed to caching the content that produces the HTML?

[DamianEdwards]

The case is every time I show it to people, they ask for it. StackOverflow is well known for caching rendered HTML in Redis. He user scenario is sound. Don't really care how you implement it, as long as it's in the one helper.

[Eilon]

Moving out of backlog. @sebastienros please post a proposal to discuss.