Noda Time is a work in progress. It has various limitations, some of which we'd obviously like to remove over time. Here's a list of some aspects we'd like to improve; see the issues list for others.
We also have a roadmap of intended releases. This is always tentative, of course, but it helps to give some clarity to our decisions in terms of what to work on next.
If there's something that should be within Noda Time's scope, but we don't support it yet, then please either raise an issue or post on the mailing list.
The .NET API provided for portable class libraries is more limited than the
full desktop version. Currently this provides relatively few challenges for
Noda Time, with one significant exception:
TimeZoneInfo. While we are able
to detect the local time zone's TZDB equivalent through
instead of its ID (as we would do normally), we can't fetch arbitrary time zones
by ID, nor can we ask for the adjustment rules for a particular time zone.
The upshot of this is that we can't currently support
on the PCL version of Noda Time.
Additionally, the PCL doesn't support .NET resource files as fully as the desktop framework; in particular, it doesn't allow you to retrieve non-string resources. This has provoked a change from the previous resource-based format used for TZDB, to a stream-based format, which is now the default. For most users this will be a no-op change, but it does affect how you build and use a custom version of TZDB.
Currently there's no way of parsing or formatting a
OffsetDateTime, partly due to
the difficulty of representing time zones or offsets. While this is
the primary deficiency of our current text support, it's not the
only one. Some other types lack flexible formatting, and we may want
to optimize further at some point too.
We'd like to be able to bundle appropriate patterns (and other internationalizable materials) within Noda Time while keeping it as a single DLL. (Satellite DLLs are fine for some scenarios, but messy in others.) Additionally we'd like to allow these resources to be augmented or replaced by the caller at execution time, to allow hot-fixes for cultures which we don't support as well as we might.
CLDR provides useful information about time zones such as a canonical ID and user-friendly representations (countries and sample cities). We'd also like to make it clearer when one zoneinfo time zone is an alias for another.
There will probably always be more calendars we could support. The highest priority is probably an adapter for the BCL calendars.
As noted in the arithmetic guide, arithmetic using
Period is pretty simplistic. We may
want something smarter, probably to go alongside the "dumb but
predictable" existing logic. This will definitely be driven by real
user requirements though - it would be far too easy to speculate.