Sounds a lot like if 'early access' was a thing, and duke nukem forever released a half baked first level that was difficult to use, and didn't really amount to anything beyond being able to point to a 'release'
Except that this is a protocol, not a product. And there are multiple teams building implementations.
HTTP 1.0 didn't have some pretty fundamental features like keep-alive.
But regardless of whether or not LN even ever works well, there will be something that does. Maybe not bitcoin, but a lot of smart people are trying a lot of different things. Just cause pong sucked doesn't mean there can't be great games.
Unsolvable like internet/BGP routing? I guess internet doesn't work either.
Tech evolves, most teams working on LN are working on getting the basic cases to a rock-solid state before doing more R&D on different routing strategies although there are already proposed alternatives on the mailing list.
No, unsolvable like routing payments from A to B requires finding routes from A to B that are well enough funded to allow the payment to go through, but with no knowledge of the funding states of the channels (as this would be a privacy leak)
The Byzantine Generals problem was "unsolvable" until Satoshi Nakamoto sidestepped it to make Bitcoin work. You don't need perfect solutions to have practical applications.
It's such a stupid argument against LN. I can make the same argument that you connecting to Hacker News is unsolveable (there is no mathematical proof that you will be able to connect to HN, because various gateways could drop offline). What does this matter in practice? You can still get here most of the time.
In LN, if the first route from A to B does not have the capacity to make the payment, an alternative route is tried. Many routes can be attempted, because they're usually sub-second negotiations between peers. If it happens that there is no route between A and B with sufficient htlc capacity that can be found in a timely manner, then you might be out of luck. Can always fall-back to an on-chain transaction. A LN invoice already supports a fallback bitcoin address. As the network grows this will become less and less likely because there will be an enormous number of routes to attempt for any payment. Chances of not making a payment will usually indicate your own or the payment destination nodes are the ones lacking capacity to make the payment.
Proof that I will be able to connect to HN is a lot different and less important than payments. It's also far from the only problem LN has.
If the network grows much, people creating and closing channels (two on-chain transactions) will start to gum up the system. If they don't close channels then they need to keep a node online or face various forms of attack. If they delegate that task to someone else, we've got yet another party to trust in a system that's already a failure when it comes to trustlessness. Even before all this, to make it work properly, vast quantities of BTC will need to be committed to the effort. The LN seems to me to be a joke.
And beyond that, well, what's this story about? Nobody really transacting in bitcoin anyway.
