lix/doc/manual/bugs.xml
Eelco Dolstra 5396304c73 * Use setre[ug]id() instead of setres[ug]id(), since the former is
more common than the latter (which exists only on Linux and
  FreeBSD).  We don't really care about dropping the saved IDs since
  there apparently is no way to quiry them in any case, so it can't
  influence the build (unlike the effective IDs which are checked by
  Perl for instance).
2004-09-09 15:55:31 +00:00

115 lines
4.6 KiB
XML

<appendix><title>Bugs / To-Do</title>
<itemizedlist>
<listitem>
<para>
The man-pages generated from the DocBook documentation are ugly.
</para>
</listitem>
<listitem>
<para>
Generations properly form a tree. E.g., if after switching to
generation 39, we perform an installation action, a generation
43 is created which is a descendant of 39, not 42. So a
rollback from 43 ought to go back to 39. This is not
currently implemented; generations form a linear sequence.
</para>
</listitem>
<listitem>
<para>
Unify the concepts of successors and substitutes into a
general notion of <emphasis>equivalent expressions</emphasis>.
Expressions are equivalent if they have the same target paths
with the same identifiers. However, even though they are
functionally equivalent, they may differ stronly with respect
to their <emphasis>performance characteristics</emphasis>.
For example, realising a closure expression is more efficient
that realising the derivation expression from which it was
produced. On the other hand, distributing sources may be more
efficient (storage- or bandwidth-wise) than distributing
binaries. So we need to be able to attach weigths or
priorities or performance annotations to expressions; Nix can
then choose the most efficient expression dependent on the
context.
</para>
</listitem>
<listitem>
<para>
<emphasis>Build management.</emphasis> In principle it is already
possible to do build management using Nix (by writing builders that
perform appropriate build steps), but the Nix expression language is
not yet powerful enough to make this pleasant (?). The language should
be extended with features from the <ulink
url='http://www.cs.uu.nl/~eelco/maak/'>Maak build manager</ulink>.
Another interesting idea is to write a <command>make</command>
implementation that uses Nix as a back-end to support <ulink
url='http://www.research.att.com/~bs/bs_faq.html#legacy'>legacy</ulink>
build files.
</para>
</listitem>
<listitem>
<para>
The current garbage collector is a hack. It should be
integrated into <command>nix-store</command>. It should
delete derivations in an order determined by topologically
sorting derivations under the points-to relation. This
ensures that no store paths ever exist that point to
non-existant store paths.
</para>
</listitem>
<listitem>
<para>
There are race conditions between the garbage collector and
other Nix tools. For instance, when we run
<command>nix-env</command> to build and install a derivation
and run the garbage collector at the same time, the garbage
collector may kick in exactly between the build and
installation steps, i.e., before the newly built derivation
has become reachable from a root of the garbage collector.
</para>
<para>
One solution would be for these programs to properly register
temporary roots for the collector. Another would be to use
stop-the-world garbage collection: if any tool is running, the
garbage collector blocks, and vice versa. These solutions do
not solve the situation where multiple tools are involved,
e.g.,
<screen>
$ nix-store -r $(nix-instantiate foo.nix)</screen>
since even if <command>nix-instantiate</command> where to
register a temporary root, it would be released by the time
<command>nix-store</command> is started. A solution would be
to write the intermediate value to a file that is used as a
root to the collector, e.g.,
<screen>
$ nix-instantiate foo.nix > /nix/var/nix/roots/bla
$ nix-store -r $(cat /nix/var/nix/roots/bla)</screen>
</para>
</listitem>
<listitem><para>For security, <command>nix-push</command> manifests
should be digitally signed, and <command>nix-pull</command> should
verify the signatures. The actual NAR archives in the cache do not
need to be signed, since the manifest contains cryptographic hashes of
these files (and <filename>fetchurl.nix</filename> checks
them).</para></listitem>
<listitem><para>We should switch away from MD5, since it has been
cracked. We don't currently depend very much on the
collision-resistance of MD5, but we will once we start sharing build
results between users.</para></listitem>
</itemizedlist>
</appendix>