Features missing from the JavaScript wrapper

[swcurran]

A couple of key features that seem to impact performance are in the Python wrapper, but not in the JavaScript wrapper. Consider adding those features.

• When opening a pool, the Python wrapper has the option of doing a "refresh" of the pool, so that the (often incomplete) list of Pool Ledger transactions in the Genesis File. This can result in a performance problem if some of the nodes in the genesis file are no longer active.
  • This change may not be needed depending on what happens with #240 .
  • This change may not be needed if the caller (e.g. Aries Framework JavaScript) immediately calls the "refresh()" method immediately after opening a pool, and keeps an updated list of transactions (vs. continuing to use the genesis_file transactions).
• When doing a read from the ledger, the Python wrapper has the options to specify the number of read nodes to use, while the JavaScript wrapper does not. It may be good to enable that option.

[swcurran]

@andrewwhitehead @cvarjao @jleach @WadeBarnes -- Did I get that right?

[andrewwhitehead]

In the Python wrapper, the open_pool method actually takes a 'no_refresh' parameter with the assumption that it should always perform a refresh before returning.

I couldn't find it before, but it appears that the nodejs and react-native wrappers do have a 'setConfig' method that can be used to set the pool configuration globally. This should be called before creating any pool instances, if it needs to be adjusted. For speeding up the initial setup I would explore setting the request_read_nodes parameter to 1 (default is 2) which defines the minimum number of nodes to send each request to. Along with this change it might be beneficial to reduce ack_timeout from 20 seconds; this is the time it will wait before dispatching another request, when the first node is slow to respond. The full set of options is here (could also use documenting elsewhere): https://github.com/hyperledger/indy-vdr/blob/323f28b547424b7f7559330eb18eed1d23ed34c2/libindy_vdr/src/config/types.rs#L8 When connecting to multiple ledgers at startup I would also make sure that those connections are performed in parallel.

[swcurran]

Good stuff! Questions:

ACA-Py leaves the “no_refresh” parameter as is, correct? Is the “refresh” that results from checking that parameter get executed by the Python wrapper, or within Indy VDR itself?

What do you suggestion lowering ack_timeout to?

I’ll get a PR in to the documentation here, and I’ll put an issue into AFJ for the changes that might be/are needed.

[swcurran]

Closing this issue as PR #242 has been completed where the Global Defaults were adjusted.

AFJ and other libraries are still able to update the defaults, but at least the defaults are improved.