Dear Osmocom community,
A little over a month ago I announced availability of this new MGW:
https://osmocom.org/projects/retro-gsm/wiki/ThemWi_E1_Abis_MGW
So far only two people responded to this news: Csaba has been testing this new MGW on his Nokia MetroSite (he and I are working toward adding AMR support), while Keith took only a passing notice - if I understood him correctly (please forgive me if I misunderstood!), it looks like Rhizomatica/TIC no longer operate GSM networks with Ericsson RBS gear, hence no more need for improved E1 MGW from their side.
However, I am getting a sense that the larger Osmocom community may be disappointed regarding my choice to develop this improved E1 Abis MGW as a "rogue" project external to Osmocom, rather than in the form of patches to osmo-mgw. Therefore, I have given some thought to how it might be possible to merge this new E1 MGW into mainline Osmocom.
Here are my current thoughts:
1) Integrating improved functionality of tw-e1abis-mgw into osmo-mgw in a fully unified manner, full coexistence of all existing osmo-mgw functionality plus all tw-e1abis-mgw functionality within a single program/process, running on one MGCP port and providing both rtpbridge and E1 endpoints, is beyond my ability, or at least beyond what I can do using only my own volunteer time. osmo-mgw code base is extremely complex, and as much as I dread to say it, insanely and needlessly so - it is a Gordian knot, and I tend to approach those the way Alexander did...
2) As an intermediate compromise between the too-difficult goal of point 1 above and the current status quo of tw-e1abis-mgw existing externally to Osmocom, how would Osmocom community feel about a hypothetical osmo-e1-mgw program? Suppose I were to produce patches that add osmo-e1-mgw as a new program to osmo-mgw git source tree, such that osmo-mgw and osmo-e1-mgw binaries would be built side by side in the same source tree - any interest? The actual code bases compiling into these two binaries would still be mostly separate, just the autotools infra of osmo-mgw git tree would be shared, but the new osmo-e1-mgw program would carry Osmocom branding, and it would live in a mainline Osmocom repository where people other than me can shape its future direction.
Please note that I don't need this integration for myself: for our own use on our to-be-built network, we (A2GC) are quite happy running the hybrid software stack where some components are built from Osmocom source repositories while other components are built from ThemWi repos. Instead I am making this offer to Osmocom community for social reasons: I wish to continue as an active member of this community, if possible, and fork-like activities are certainly _not_ the way to go, socially.
I am also making this integration offer to the community as food for thought, without any urgency. If the greater Osmocom community does wish to see tw-e1abis-mgw turned into something like osmo-e1-mgw, integrated into mainline Osmocom, it will probably be another few months before I can start submitting patches in this direction. The reason for the delay is that tw-e1abis-mgw is not quite finished at the present moment: two big missing features are AMR support and UL-to-DL transform (aka TFO transform) for EFR. I would like to fill these two functional gaps in the new MGW while staying in my own tw-e1abis-mgw source tree, before beginning the process of merging into Osmocom mainline - hence the time delay prognosis.
If Osmocom community does wish to see the improved E1 MGW as part of mainline Osmocom code base, I would need an explicit "green light" go-ahead before I start preparing any patches. The patch process would need to proceed as follows:
1) I will need to prepare a patch for libosmocore, adding a header file with definitions from 3GPP TS 48.103 Table 5.4.2.2.1. Right now ThemWi software, including tw-e1abis-mgw, uses <themwi/common/aoip_rtp_pt.h> header file - but of course the latter won't be available in pure Osmocom environment without ThemWi dependencies.
2) There will need to be a whole bunch of patches to libosmo-abis.git (libosmotrau portion thereof), adding various functions which I currently collect in libtwtrau and which will be used by the upcoming AMR-capable version of tw-e1abis-mgw.
3) Only after 1 and 2 above are merged, then I'll able to produce a big patch for osmo-mgw.git, adding osmo-e1-mgw (or whatever other name the community comes up with) as a new program.
In any case, in the immediate short term I am going to continue improving tw-e1abis-mgw where it lives currently, in ThemWi area of Osmocom Gitea. Once AMR support and UL-to-DL transform for EFR are there, I will make another post here. At that point if the greater community would like to see this work merged, I will need feedback in this ML thread.
GSM/2G Forever, Mother Mychaela
I can't tell the future at the moment; I updated the OS on my crystal ball a few months ago and it still hasn't rebooted yet.
It's quite possible that at some point, somebody might want more 2G systems, Ericsson or not. Run by Telecomunicaciones Indígenas Comunitarias, or not, - and/or 2G/LTE or even whateverG else is around then.
Improved MGW support for E1 is of course, er.. an improvement. Lack of AMR there would be on my list of laments for what I was unable to get done (yet), along with all those other things to do with audio quality with AMR + DTX and all those weird things to do with RTP streams and audio on the radio interface, that Mychaela seems to be quite passionate about.
Also listed would be, multiple radio support in osmo-bsc on the DUG, and quite a big lament; - Lack of dual mode radio support for a 2G/4G combination on one RUS/sector. Anyway, no, I don't have any use case for it at the moment, and my motivations for spending hours coding 2G had diminished somewhat along with TIC-AC's changes in direction, and the pandemic... and the war.. and the genocide.... Besides, after all these years, well.. the femto BTS (osmo-bts) based systems are working now very stable, although rough working conditions has taken its toll on a lot of that radio equipment.
And, it seems the the flood of Ericsson equipment onto the second hand market that was once anticipated has yet to materialise.
Anyway, all I really wanted to say was that I don't see any evidence that osmocom community considers anything "rogue" and if that was interpreted from a comment* of mine, it should not have been.
* now as I write realising that comment was offlist.
On Tue, Oct 07, 2025 at 08:14:39PM -0800, Mychaela Falconia wrote:
So far only two people responded to this news: Csaba has been testing
...
while Keith took only a passing notice
...
him correctly (please forgive me if I misunderstood!), it looks like Rhizomatica/TIC no longer operate GSM networks with Ericsson RBS gear, hence no more need for improved E1 MGW from their side.
However, I am getting a sense that the larger Osmocom community may be disappointed
Hi Mychaela,
I would like to reflect that the above paragraphs are very uncomfortable for me to read.
You do have improved your communication dramatically from the first time we discussed mailing list netiquette, thank you for that. However, I feel that this is a subtle gem, and I do see the need to give some feedback.
The points in particular:
Be aware that this is a public forum:
- I would prefer if you let the other developers step into a public thread *if* they choose to, and not publicly call out names when you don't have to.
- Please do not infer, hypothesize, let alone leak details about how someone else runs their site. It is their choice to disclose those details publicly -- or not! It is inappropriate to poke for details there.
Reading that you "sense" that all of us are disappointed by your work reads like a provocative overstep. We are not close enough that I could in any way be disappointed.
Let's keep a neutral tone on a technical mailing list: Please keep your assumptions on low profile, and please invite instead of poking and prodding.
Thank you!
~N
-
Keith -
Mychaela Falconia -
Neels Hofmeyr