I can see both sides here.
Ultimately though, the universal truth is... people are lazy, and will go for the most convenient option rather than the more "soverign" option, even willingly knowing the risks. The future risk of a centralised service collapse, to most people, is less scary than spending an uncomfortable amount of time understanding a difficult app setup.
Fazer is right this is a necessity, and I fully understand his passion for it... equally I am the same, but I have to set aside the deeper developments for things that people find convenient and enjoyable - first impressions mean a lot, I realise.
Eric is spot-on with the "hybrid" approach, this is the Labs way - MyPIVXWallet has greatly progressed, despite launching as a client with absolutely bare-minimum chain validation, and the worst possible privacy practices (ironic for a privacy coin)... almost 3 years later it now has block sync capabilities (thus SHIELD balances can be synced without MPW's explorers knowing the balances), and it's even possible for us to allow MPW to sync Public balances without revealing a single public key - the point is, you have to make compromises, in order to progress.
That said, I am very stuck with this proposal: I want to see how this can be put in to users' hands, and how non-technical people can enjoy trading on a layer-1 DEX without spending a full day setting up wallets, synchronising, doing DevOps and all the other time consuming tasks one needs to perform to make most layer-1 DEXes 'usable'.
Now... if this system was integrated with something like XLite, so that users could setup automatic market making, orders, etc: in the interface of a lightwallet, I would be VERY interested, not only is this technically achieveable: but the UX is sound, people WOULD use this. This is why people love "Uniswap", despite it being, in my opinion as well, a "fake" DEX. We need a real DEX with lightwallet capabilities, until the intention to build that is set, I'm afraid I cannot put too much time in to this.
- JSKitty