We have been overly pessimistic in terms of expected certificates TLV size. The spec is clear the certs should be smaller than 400 bytes, so I think this is safe (and Matter provisioning with Google passes fine).
This saves in total ~ 16K, as certs are everywhere (fabrics and sessions' nocs).
=============================
The next change to shortly follow is also a relatively low hanging fruit (stuff in the QR code generation).
The changes that will follow afterwards though are much more obtrusive (as in making the exchanges to share a single TX/RX buffer whenever possible). But exchange cleanup is due anyway (and a bit necessary so as to address retransmissions and to make sure we are behaving when the user drops an unfinished exchange in the middle of the transmission due to errors).
We have been overly pessimistic in terms of expected certificates TLV size. The spec is clear the certs should be smaller than 400 bytes, so I think this is safe (and Matter provisioning with Google passes fine).
This saves in total ~ 16K, as certs are everywhere (fabrics and sessions' nocs).
=============================
The next change to shortly follow is also a relatively low hanging fruit (stuff in the QR code generation).
The changes that will follow afterwards though are much more obtrusive (as in making the exchanges to share a single TX/RX buffer whenever possible). But exchange cleanup is due anyway (and a bit necessary so as to address retransmissions and to make sure we are behaving when the user drops an unfinished exchange in the middle of the transmission due to errors).