Pavel Emelyanov
Pavel Emelyanov
> Can't S3 files handle append-only in dma_write? That's the biggest problem. Typically output_stream + file_data_sink_impl would push 128-k buffers into S3 file in a form of `void* buf, size_t...
Yes, it works perfectly for S3. And this is how it's currently implemented in my branch, I just though that it could be nice generalization for seastar.
The goal of things in demo/ subdir is to demonstrate the usage of a particular seastar API be if file IO, TLS sockets, etc. Does this word-count demo demonstrate something...
In this case, I think the demo deserves at least 5 improvements to begin with - make filename a CLI option - add a long comment describing the counting algorithm...
Collected from i4i.4xl by @pwrobelse 
@nyh , @tchaikov this is the restart implementation without magic numbers (but more hairy) as I mentioned [here](https://github.com/scylladb/seastar/pull/1847#discussion_r1361782978). Mind taking a look. Patches 1 through 5 are identical in both...
upd: - returned back runaway comment
@nyh , @tchaikov , please review (or re-review the #1847 one)
@tchaikov , thank you for the review. I'd also like you to share your thoughts about which restarting mechanism you like more -- this one (that replays the request once...
@nyh , reminder ping. Please tell which approach sounds more reasonable to you -- this or #1847