MSG isn't validated

[dra27]

RFC3164 4.1.3 describes the two fields of the MSG portion, but this isn't validated anywhere. In principle, the encode function should probably raise if there is no discernable alphanumeric TAG of 32 chars or fewer delimited by a non-alphanumeric character in message.

[verbosemode]

We could also replace the message field in type t by a tuple, record or even a new type to accommodate for the tag and content field. This would break the API, but afaik the only consumer of this library is logs-syslog at the moment. If @hannesm is fine with it, this might be an option to raising an exception.

[Leonidas-from-XIV]

I think this would be preferable, since exceptions tend to be rather surprising and break your code on runtime. The other possibility is to use the result type with polymorphic variants, I've also made good experiences with that.

[hannesm]

I'm happy with a (breaking) API change :)
FWIW: I like to use the result type for decode style functions (i.e. input from the network), but encode (i.e. input from trusted source, such as hardcoded parameters) I'm happy with Invalid_argument exceptions.

and I don't expect that this tag comes dynamically from untrusted sources..

[hannesm]

see #20