Network discovery estimation in release lifecycle

If ITN experience was anything to go by, the biggest hurdle for the entire ITN lifecycle was the phase between introduction to stabilisation of decentralised peer-to-peer networking.

The issues discovered ranged from tuning TCP parameters at OS level, memory optimisations within code, tuning number of connection requests made by nodes and so on.

While IOHK is very confident of possibly not hitting as many issues with Haskell implementation, the initial round of testing (in absence of network discovery) does not go with the same tone
(i.e. ofcourse, excluding testing ideal case scenario - viz, synchronised topology.json between x number of nodes spawned on similar OS, with similar devops scripts, same parameters, config, health status, etc - being monitored based on specific set of measurements)

Now, I understand that we would want the basics to be working well before going on to Network Discovery, I would expect the release cycle not dates and plan to be clear (and shareable) from IOHK about when in the lifecycle should we expect the network discovery being enabled?

I would expect this to be part of minimal viable milestone before a measurable public testnet can be worked upon. In absence of results of this critical component, I find it very difficult to measure how well (or not) the experience is.


This echoes my own thoughts quite well. Without a working networking protocol, I’m not sure you can call the system decentralized, and to my mind, that is the ultimate goal.

I would not. I would even go so far to say, it makes more sense to introduce this feature (the governor) as soon as it is ready and well tested and not to throw it on the end users (pool operators in this case) without almost no testing, like it was the case on the ITN.

Given we will go to mainnet with decentralisationParam still laveraging IOHK side and that will be case for months, we still won’t be able to speak about decentralization even if we have perfectly working P2P discovery.

I am with you on this , but to me it converts to what I said - not having public testnet until its ready, however long it takes . As we wouldnt want to introduce network breaking changes over an already populated network.

Regardless, the intent of the post is clarity from IOHK about when in lifecycle is it planned.


I see another non-technical reason as well : if the testnet goes public without p2p, I don’t see pioneers constantly changing their topologies to make space for new operators. That might provoke a crisis and another round of tiring discussions.


Agreed, it would be disastrous, and is an absolute prerequisite for mainnet to start without.

Given the recent dates it looks like it wont be in for initial public testnet, which feels like a bad step already. Hopefully everything just magically works as soon as its constructed - but we cannot live on hope when exact dates have been shared :slight_smile: