self mutable reference

[MathiasKoch]

I am a bit unsure on why this shouldn't have a mutable reference to self?

The way i see it, you would usually implement this on some sort of network client, owning access to a peripheral (unix fd, SPI, I2C, Serial etc.), but most (all?) of these require a mutable reference to be able to send data. This could be fixed by wrapping the peripheral in a RefCell, but even then there is a good chance that the client object would hold stuff like internal state etc that needs to be mutated, meaning that in my particular case, i would end up with a client object of 5 fields, all wrapped in RefCell.

Is this intended behavior, or am i implementing something wrong in general?

[thejpster]

I had intended implemetors of the crates to handle internal mutability. While in the BSD socket library, I can call open() or write() etc, without having to hold any kind of pointer or reference - it is handled internally by the OS - here, having to hold a non-mutable reference to a particular IP stack seemed like a good compromise.

Requiring a &mut ref would make operating, say, an HTTP client and an SMTP client at the same time almost impossible.

[MathiasKoch]

I think that makes sense. I will go with the internal mutability 👍

[kellerkindt]

Cross linking to #39