Re: [Haskell-cafe] Fast JSON validation - reducing allocations

David Turner writes:

On 12 May 2017 at 14:44, Ben Gamari wrote:

...

For this reason, if your program is truly embarassingly parallel, it is usually preferrable performance-wise to run multiple processes concurrently than use pure parallelism (a sad truth in a harsh reality, in my opinion).

Hmm, an interesting idea. It's not _all_ embarassingly parallel, but the bottleneck is. You're right, we could have separate front-end processes just doing the validation (indeed they could even be distributed across the network). I don't think we'll do that just yet, particularly not when 8.2's results are so promising, but it sounds a useful idea to keep in the toolbox.

Indeed. Note that Haskell is also quite good at distributed computation. StaticPointers makes it relatively easy to talk about code distributed across processes and machines. See, for instance, the Cloud Haskell suite of libraries. Cheers, - Ben