There seem to be two questions here; whether it’s good for the network, and how to decide whether to use it.
Is it good for the network?
This is a pretty complex question that we debate even inside the GasToken team.
On the one hand, you are correct; GasToken appears to have a natural smoothing effect and allows the quantity supplied to be more responsive to demand. On the other hand, for UX, it’s not clear to me that this will be a positive shift; it’s entirely possible that the bounded stable gas price in a steady state system using GasToken could be higher than what users are paying today.
There are several factors worthy of consideration.
- Influence of speculation: an efficient and liquid GasToken would imply speculation as a use case, where no direct/trustless speculation is possible on gas today (second layer speculation where users do not directly hold gas is possible through e.g. prediction markets, but GasToken is a more direct tokenization mechanism for the underlying computation). Computation in Ethereum is relatively scarce (as recent gas issues indicate), leading to the possibility of upwards pressure on gas price worsening UX for users. One analogy is the only scarce resource in EOS, RAM, representing storage, which was opened to speculation, and has seen uncertain market dynamics and hoarding as a result. It is entirely possible that the eventual outcome is continuous releases of supply-by-decree, overinflated costs to users, or an accurate and robust pricing mechanism. The latter would hint that speculation/hoarding is not a big issue. One key difference is that GasToken represents (essentially) tokenized promises of computation subsidies, rather than tokenized storage.
- Overall capacity reduction: If GasToken is used for every transaction, the overall capacity of the network will be reduced over the current market with full blocks, as GasToken incurs efficiency loss in accounting/call overhead. Miners can vote the gas limit up to somewhat compensate for this, but after this becomes stable you run into the issues in (3).
- Useless resource consumption: GasToken uses resources in the system. If the platform is otherwise optimizing for whatever total capacity it would like to provide for a given security level (in terms of computation, network, and storage), this waste means reduced total capacity or reduced total security. This could be partially mitigated through client-side optimizations, and possibly solved through a variety of hard fork changes (e.g., replacing GasToken with a simulated version, or more fundamental changes to the resource model). It could also be mitigated by disabling GasToken through removal of the refund, or by making its use less attractive by e.g., decreasing the percentage of transaction fees that is refundable (though this would neuter refunds so severely the question of “why keep them” remains obvious; one nice property of GasToken is that in a liquid form it actually provides a super nice pathway for incentivizing cleanup).
So is GasToken good for the network? For now, perhaps. In the long run, it probably isn’t the optimal solution.
Should you use it?
I think that should purely be a question of whether it is economically rational for you as a participant, not whether it is good for the network. As long term holders and investors in the ecosystem, we all have an incentive to make sure the network survives. That being said, we can’t continue to rely on or assume such altruism as the system grows. If an unpermissioned system is to survive, it must be able to tolerate economically rational behavior inside the platform. So personally, I would encourage whatever decision makes economic sense in the short to medium term.
Right now, using GasToken is clearly economically rational for short-term gas bidders such as arbitrageurs or serial ICO buyers, even in the absence of a liquid market (as mining yourself is practical here). It is also clearly economically rational for businesses who plan on doing large gas transactions reasonably into the future and want to reduce risk surrounding a Cryptokitties or F-Coin-like event. In all of these cases, we are dealing with sophisticated players who use large amounts of gas and thus have an incentive to put in the engineering effort to implement their own GasToken-like mechanism or integrate with GST. Anyone professionally trading on DEXes, sweeping contracts, doing an airdrop, etc. likely falls into this category. This isn’t just hypothetical — we are aware of users who are using GasToken-like mechanisms in their smart contracts today.
For general-purpose contract use (as in the case of Gnosis Safe), a liquid market would be required for maximum utility. So whether you should use it really depends on whether such a market arises or not. A handful of large-ish players like Gnosis would be required to make such a market truly robust. The accounting/exchange overhead should also be minimal and the per-transaction gas usage relatively high to make such use viable; the transaction calculator can let you know whether this is the case today — I am sure that you did the math and determined that it would work in the case of Gnosis Safe.
Finally, there also is a “fairness” argument: sophisticated players are already taking advantage of GasToken/GasToken-like mechanisms. It would be nice if the benefits were also made available to the general public. Gnosis Safe might be a good starting point for this.
From a research perspective, there is a massive silver lining: we can learn a lot from GasToken deployment that will highlight fundamental issues in current resource models and help us design next-generation mechanisms (as Lorenz pointed out this is the aim of the overarching project, Project Chicago).
All the usual disclaimers on the GasToken site apply; there is substantial risk, especially in holding GST1/GST2 long term.
(Thanks to GST team members Lorenz, Florian, and Ari for providing input/contributions on this post.)