11 |
This would unify paxutils with tar. |
This would unify paxutils with tar. |
12 |
|
|
13 |
* Remove command-line incompatibilities between GNU tar and UNIX tar |
* Remove command-line incompatibilities between GNU tar and UNIX tar |
14 |
as specified by UNIX98. The main problems are: |
as specified by UNIX98. The main problem is: |
15 |
|
|
16 |
l GNU tar doesn't cross filesystem boundaries. |
l GNU tar doesn't cross filesystem boundaries. |
17 |
UNIX98 tar warns if all links cannot be resolved. |
UNIX98 tar warns if all links cannot be resolved. |
18 |
(GNU tar has no similar option.) |
(GNU tar --check-links option) |
19 |
|
|
20 |
o GNU tar says output old format. |
Perhaps we could announce a phase-in period where "l" changes in semantics. |
21 |
UNIX98 tar says don't chown files after extracting |
In the meanwhile we could make the "l" semantics to be determined by the |
22 |
(GNU tar uses "--no-same-owner" for this). |
value of POSIXLY_CORRECT variable. |
|
|
|
|
Many people rely on GNU tar's "l" and "o" so we can't abruptly change |
|
|
GNU tar to be compatible with UNIX98 tar. |
|
|
|
|
|
However, "o" is relatively easy: we can extend GNU tar so that "o" is |
|
|
a synonym for --no-same-owner when extracting, keeping "o"'s current |
|
|
semantics when creating. This is a bit of a hack, but it shouldn't |
|
|
break things. |
|
|
|
|
|
"l" will be harder, since it's a clear incompatibility. Perhaps we |
|
|
could announce a phase-in period where "l" changes in semantics. |
|
|
However, the first step should be to support the UNIX98 semantics, |
|
|
using a long option, without changing the meaning of "l" for now. That |
|
|
will give people a migration strategy. |
|
23 |
|
|
24 |
* Interoperate better with Joerg Schilling's star implementation. |
* Interoperate better with Joerg Schilling's star implementation. |
25 |
|
|