Da service backend #381
Da service backend #381
Conversation
danielSanchezQ
force-pushed
the
da-service-backend
branch
from
164429d
to
87d4d99
Compare
Codecov ReportPatch has no changes to coverable lines. ❗ Your organization needs to install the Codecov GitHub app to enable full functionality. 📢 Thoughts on this report? Let us know!. |
| pub type BlobHasher<T> = fn(&T) -> <T as Blob>::Hash; | ||
|
|
||
| pub trait Blob { | ||
| const HASHER: BlobHasher<Self>; | ||
| type Hash: Hash + Eq + Clone; | ||
| fn hash(&self) -> Self::Hash { | ||
| Self::HASHER(self) | ||
| } | ||
| fn as_bytes(&self) -> Bytes; | ||
| } |
There was a problem hiding this comment.
This may be expanded later if needed when we know more about how would it looks.
|
|
||
| pub fn pending_blobs(&self) -> Box<dyn Iterator<Item = B> + Send> { | ||
| // bypass lifetime | ||
| let blobs: Vec<_> = self.0.iter().map(|t| t.1).collect(); |
There was a problem hiding this comment.
moka clones by default.
| // can we leverage this to evict really old blobs? | ||
| .time_to_idle(evicting_period) |
There was a problem hiding this comment.
As discussed in https://github.com/logos-co/nomos-node/pull/376/files#r1317083701, the expectation is that the outer service retrieves pending blobs from this service, does something with those blobs, and after that, requests removing those blobs? If so, it seems it's okay to use time_to_idle. This will reset eviction timers whenever pending_blobs() is called, but I guess it's okay if it wouldn't take too long for the outer service to process and remove blobs.
There was a problem hiding this comment.
Yes. So the idea here is that. Either a blob is pending for the whole life until is added into a block. Or is removed. (that is why is using the same evicting period as the time to live). But it would make sense that if it is not included it dies as well after some time.
No description provided.