Send from ROM memory

[TomzBench]

I've read the tutorial here about sending large amounts of data and find it a little lacking in the scenario where memory is stored in ROM. The premise of the huge response tutorial is that embedded devices should chunk up requests into smaller chunks, because embedded devices have limited RAM.

However, it is possible to send large amount of data from ROM. We can bypass any mem copies into the io buffers and simply maintain state of how much is left to send by tracking sent bytes on each call from poll, and simply increase a index tracking a pointer in memory in ROM.

There are a few use cases for this scenario.

1. We have an embedded system that logs events into it's persistent storage. A client would like to read this log history from the persistent storage.

2. You can also serve packed web images from const memory instead of copying into io buffers.

3. probably more!

[scaprile]

Perhaps you should do that by storing your data in an embedded file system. The concept of that tutorial is not about constant data.
https://mongoose.ws/documentation/tutorials/core/embedded-filesystem/

I don't envision we might pursue anything other than that, as in our measurements the copy overhead is negligible.
However, @cpq WDYT ?

[TomzBench]

Perhaps you should do that by storing your data in an embedded file system.

This might work. I will have to review.

as in our measurements the copy overhead is negligible

Sure - the zero copy is just a nice side affect if you have the const memory feature is all. Also - sending from const memory is more simpler than managing buffers. So in terms of maintenance complexity - it might be nice to add.

[TomzBench]

I think using the filesystem virtualization layer, i can model my ROM as a file system with a single file, then the client can request this "file". And i think it would work for my ROM use case. Perhaps can close then.