Consider Load-Link Store-Conditional abstraction

[jzabroski]

Rather than locking your type buffer, consider implementing a Load-Link Store-Conditional abstraction. Rather than worry about the state, just retry the whole chain if Store-Conditional fails.

The main benefit I see is it would potentially uncomplicate the code a lot. I'm not sure if it would be any faster, but it's conceivable it would.

[jzabroski]

If you're interested, I know my friend Sandro Magi has written a LLSC abstraction in his Sasa NuGet package in the Sasa.Concurrency namespace. That would probably lower the bar for prototyping getting rid of the lock requirements.

[gafter]

@jzabroski Because the hash table has to be expended by doubling the size of the array and rebucketing the data, I don't think that technique works well. In any case, we only lock on writes, not on reads.

[gafter]

I should also mention that the implementation is intended to be suitable for use in the core library. Taking advantage of a higher-level library would undermine that.