David Frese

> Can't understand why connect is calling multiple times, if :jamdb_oracle.start_link fails. I would say it is not supposed to fail. Postgrex fails when essential config options are missing, but...

> [fixed](https://github.com/erlangbureau/jamdb_oracle/commit/cdda5108a653791be0b33789e82442e437d3adce) !? Uhh, I would expect a lot of other problems following from that. I think the solution on the `stage` branch is what DBConnection expects from implementations. Also...

Found a little bug on the stable branch: In `query` the last case should be ```elixir {:error, err, conn} -> {:disconnect, err, %{s | conn: conn}} ``` instead of ```...

I also had some weird occurrences of errors, where a jamdb_oracle functions tried something on values that seemed to represent messages from other processes (like a Phoenix.Socket message). I looked...

> I also had some weird occurrences of errors, where a jamdb_oracle functions tried something on values that seemed to represent messages from other processes (like a Phoenix.Socket message). >...

> Packet header unfortunately doesn't contain PacketSize for long packets > 8k PacketSize is always

> 0 means that all available bytes are returned. All bytes of one packet. After each data packet of maximal size (or marker packet) recv(read_timeout) receives next packet or aborts...

That kind of helped. But now it actually waits for the full read_timeout in normal operations! I.e. it gets reaaaally slow with a read_timeout of 500ms; even the connect fails...

> Try add parameter socket_options with recbuf `parameters: [... read_timeout: 500, socket_options: [{:recbuf, 8192}] ...] ` If sdu=8192 (default) then recbuf=8192 or you can use other values 4096,2048,... Why? I...

Hi, even with my changes (not trying to decode after a socket timeout), I get sporadic :bad_match errors in production. They come in two forms, and start like ```elixir {:badmatch,...