WIP fix: yet another prefix lossy subscription sketch #5418
Conversation
erights
force-pushed
the
markm-prefix-lossy
branch
from
0433520
to
cab0a23
Compare
erights
requested review from
michaelfig,
dtribble,
turadg,
Chris-Hibbert and
gibson042
There was a problem hiding this comment.
I think I see how the introduction of the updateCount can help satisfy the "make a client-side notifier for a subscription" problem.
The reason why the implementation in #5403 handles configurable subscription history is that the other "prefix lossy" subscriptions are essentially the same as historyLimit=1, but I see legitimate uses of historyLimit=Infinity (current subscription implementation) and historyLimit=0 (start with the next published value). I want to be sure we are agreed upon breaking compatibility and restricting existing flexibility before we remove support for all subscriptions except historyLimit=1.
Having only "prefix lossy" subscriptions available to me means that some of my uses will require more bookkeeping on the client-side to consume as intended, and also while necessarily consuming more memory on the server-side.
erights
force-pushed
the
markm-prefix-lossy
branch
from
79a6311
to
3369b16
Compare
erights
force-pushed
the
markm-prefix-lossy
branch
from
3369b16
to
3a73b03
Compare
| * @typedef {object} SharableSubscription | ||
| * @property {() => ERef<SubscriptionInternals<T>>} getSharableSubscriptionInternals | ||
| * Used to replicate the multicast values at other sites. To manually create a | ||
| * local representative of a Subscription, do | ||
| * ```js | ||
| * localIterable = | ||
| * makeSubscription(E(remoteIterable).getSharableSubscriptionInternals()); | ||
| * ``` | ||
| * The resulting `localIterable` also supports such remote use, and | ||
| * will return access to the same representation. | ||
| * @property {GetIterationSince<T>} getIterationSince | ||
| * Internally used to get the "current" SharableSubscriptionInternals | ||
| * in order to make a subscription iterator that starts there. | ||
| * The answer is "current" in that it was accurate at some moment between | ||
| * when you asked and when you got the answer. |
There was a problem hiding this comment.
Bookmarking this code so that I can refer to it.
|
issue is closed #5413 |
Based on @dtribble 's at https://gist.github.com/dtribble/4e7cc65fb0fa8aab5a4069e80d915454