Quote from: iCEBREAKER on February 16, 2017, 09:29:53 PM

Quote from: IadixDev on February 16, 2017, 08:04:50 PM

Quote from: iCEBREAKER on February 14, 2017, 05:44:14 AM

A Lightning type payment channel is simply a write cache for the blockchain.

Do you know how much the write cache improves a hard drive's performance? Try turning yours off and find out.

Why don't you moan about the presence of write caches on hard drives, and spew FUD about how output operations to the write cache aren't real "on disk" transactions! Cheesy

You know it's not real "on disk" operations when the computer crash or power short age and the filesystem end up corrupted :p otherwise you would just keep all the data on ram disk and just write on the HD on shut down Wink

how safe would that really look Wink

Write cache aren't not all that fud free Wink

Now with journalized filesystem it mitigiate the possible damage, but having reliable decentralized journalizing on LN would mostly ruin the concept. And in this context having only private journalizing is not changing much.

You'd have a point, if and only if

-Bitcoin's filesystem (The Blockchain) could be shut down, crash, or suffer from power outages (it can't)
-LN didn't already journalize when opening channels (it does)
-LN used private journalizing (it doesn't)

Well as far as i understand, LN channels can be somehow shut down, via certain glitches or other, and in that case, what would remain of the operation made in that channel ?

And what validity would this journalizing have regarding on chain state if there are difference at the end ?

The problem with caching is not about HD crash, but if the controller is stopped before the cache is a actually wrote, even if the hard drive works well the data is still lost.

Just to push analogy with cache to show certain caveeat, with smp system and cpu cache, there are certain case when the memory is shared with other chips with dma, or virtual pagination, in system with high concurency on the data, cpu cache can become "out of date", even with sse2 there are certain thing to help dealing with this, but as far as i know, most os disable caching on certain shared memory because of all the issues with cache, and instruction reordering etc When having access to up to date data in concurent system is more important than fast access to potentially out of date data.

If LN is to be seen as a cache system , it doesn't look like they are taking all the precautions for it to be really safe.

Cache are easily safe when all the write access to the data are made throught the same interface doing the caching, which is not the case with bitcoin & LN.

With hard drive it works because all the access goes throught the same controller doing the caching.

But anyway as LN locks the bitcoin on the main chain, it's not even really a true cache system, because the principle of a cache system is to fasten multiple access on the same data, as the bitcoin are locked, the channel have exclusive access to it, and so it's not really to be seen as a true system of blockchain caching.