aboutsummaryrefslogtreecommitdiff
path: root/issues/server-configuration.md
blob: dab70b98507c320e2f1b23b407efdd1d68056aaa (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
# Server configuration
Reported by: ln5

All server configuration is done via command-line arguments.

This is good for configuration settings which last through the lifetime of an
invocation of a log instance, i.e., from launch to Ctrl-C.  Examples:
- `--http_endpoint`
- `--key`.

It is less good for settings that change over time.  Examples:
- `--witnesses`

Reading a configuration file at start and when receiving, say, SIGHUP, is an
alternative.

Implementing a "control port", typically a TCP endpoint, where an administrator
can "program" the log instance is another alternative. Such an interface can
also be used for diagnostics.

We also need to add better documentation on how to run and configure
sigsum-log-go.  There is a rough start in cmd/sigsum-log-go/README.md.  It
assumes a little bit too much of the reader, and does not document everything
that is relevant.  For example, configuration of sharding is not documented.