I appreciate Unruggable’s agreeing to take some of the feedback from delegates into account, and to adjust their proposal. Like @Leon and @jefflau.eth, I believe that engaging an external provider like dRPC to host the gateways is likely the most effective way forward.
I’ve worked in SRE (AKA DevOps) at Google, and I can say from experience that running world class production grade infrastructure is much harder than it seems; anyone thinking of operating gateways at this scale - and I include both Labs and Unruggable - would be well placed to consider outsourcing the actual operation to a talented partner who can offer contractual SLAs and is already running high availability nodes for the chains concerned. It will be difficult to beat the combination of latency and uptime that they can offer.
Assuming Unruggable are prepared to give this option serious consideration, I’d be prepared to support this proposal on behalf of myself and Labs, provided it were adjusted to be part of the streaming provider program. This could be achieved either via interim funding until the program is renewed next year, or by passing a separate proposal approving the funding and specifying that it will be grandfathered in to the next iteration, counting as already approved and allocated from its budget.
If we are correct about third-party providers being a more affordable way to run production infrastructure, I have no doubt Unruggable will put the budget thus freed up to good use, and demonstrate as much to the DAO.
I continue to have serious concerns with bypassing the streaming provider program; if the program is not fit for purpose, we should be fixing it rather than bypassing it, and while I believe Unruggable are more than capable of delivering what they promise, I am concerned about the precedent that an easy bypass of the streaming provider program sets for both the DAO’s budget and for future requests. I cannot support the proposal if it continues to bypass the mechanisms we have put in place for ongoing funding of ENS public goods.