In private conversation, Holger mentioned EC-GSM-IoT to me, and I had to dig a bit into it. It was introduced in Release 13, but if you do a web search for it, you find surprisingly little information beyond press releases with absolutely zero information content and no "further reading".
The primary reason for this seems to be that the feature was called EC-EGPRS until the very late stages, when it was renamed for - believe it or not - marketing reasons.
So when searching for the right term, you actually find specification references and change requests in the 3GPP document archives.
I tried to get a very brief overview, and from what I could find, it is centered around GERAN extension in the following ways:
EC-EGPRS goal: Improve coverage by 20dB
New single-burst coding schemes
Blind Physical Layer Repetitions where bursts are repeated up to 28 times without feedback from remote end
transmitter maintains phase coherency
receiver uses processing gain (like incremental redundancy?)
New logical channel types (EC-BCCH, EC-PCH, EC-AGC, EC-RACH, ...)
New RLC/MAC layer messages for the EC-PDCH communication
Power Efficient Operation (PEO)
Introduction of eDRX (extended DRX) to allow for PCH listening intervals from minutes up to a hour
Relaxed Idle Mode: Important to camp on a cell, not best cell. Reduces neighbor cell monitoring requirements
In terms of required modifications to an existing GSM/EDGE implementation, there will be (at least):
changes to the PHY layer regarding new coding schemes, logical channels and burst scheduling / re-transmissions
changes to the RLC/MAC layer in the PCU to implement the new EC specific message types and procedures
changes to the BTS and BSC in terms of paging in eDRX
In case you're interested in more pointers on technical details, check out the links provided at https://osmocom.org/issues/1780
It remains to be seen how widely this will be adopted. Rolling this cange out on moderm base station hardware seems technicalyl simple - but it remains to be seen how many equipment makers implement it, and at what cost to the operators. But I think the key issue is whether or not the baseband chipset makers (Intel, Qualcomm, Mediatek, ...) will implement it anytime soon on the device side.
There are no plans on implementing any of this in the Osmocom stack as of now,but in case anyone was interested in working on this, feel free to contact us on the email@example.com mailing list.