| 1 |
- parameter/setting default from manifest |
| 2 |
==> BRANCH[feature_default_parameters], |
| 3 |
==> PERSON[Steven or Nico] |
| 4 |
==> PROPOSAL(1) |
| 5 |
- current bug |
| 6 |
- proposal 1: parameter/default/$name (for optional ones) |
| 7 |
- new way |
| 8 |
- catches --state absent|present |
| 9 |
- needs changes of types |
| 10 |
- also possible for explorer |
| 11 |
- support for it in core? |
| 12 |
- handling of ${o} $o "$o" ? |
| 13 |
- handling which variables? |
| 14 |
- introduction of "templating language" |
| 15 |
- aka macros |
| 16 |
- possible problems: |
| 17 |
- inconsistency |
| 18 |
- redoing shell functionality |
| 19 |
- raising expectations for more templating from users |
| 20 |
- possible benefit |
| 21 |
- no need for eval |
| 22 |
- once in core, not everytime in type |
| 23 |
- OTOH: one extra word. |
| 24 |
- a=$(cat $__object/parameter/name) vs. $(eval $(cat $__object/parameter/name)) |
| 25 |
- only possible for static defaults |
| 26 |
- --name overrides name not possible vs. object_id |
| 27 |
- Is this the only case???? |
| 28 |
- if yes: don't care. |
| 29 |
- possible solution: |
| 30 |
- echo '/$__object_id' > typename/parameter/default/name |
| 31 |
- eval $(cat $__object/parameter/name) |
| 32 |
- probably allows code injection |
| 33 |
- is possible anyway??? |
| 34 |
- $(cat /etc/shadow) |
| 35 |
- other eval side effects??? |
| 36 |
- none: go for it |
| 37 |
- some: have headache |
| 38 |
- many: don't do |
| 39 |
- proposal 2: 2 dbs (user input vs. stuff changable by type) |
| 40 |
- explicit 2nd db [parameter_user and parameter/] |
| 41 |
- not very clean (both agreed) |
| 42 |
- proposal 3: parameter are read-only |
| 43 |
- breaks current types (in core probably elsewhere) |
| 44 |
- can't enforce, but user is on his own => breaks, her problem |
| 45 |
+ clean seperation between core and type (nico) |
| 46 |
- parameter belongs to type not core (steven) |
| 47 |
- proposal 4: core ignores changes in parameter/* of object |
| 48 |
- implicit 2nd db [see automagic below] |
| 49 |
- steven+++ |
| 50 |
- does not work with divergent emulator not being in core |
| 51 |
- because emulators primary db __is__ fs. |
| 52 |
|
| 53 |
1 manifest: |
| 54 |
|
| 55 |
__foo bar == emulator |
| 56 |
echo present > $__global/object/__foo/bar/parameter/state |
| 57 |
|
| 58 |
# fails |
| 59 |
__foo bar == emulator |
| 60 |
|
| 61 |
! automagic / filesystem |
| 62 |
! fsproperty: |
| 63 |
- kill, write explicitly to disk |
| 64 |
==> BRANCH[cleanup_fsproperty] |
| 65 |
==> PERSON[Steven] |
| 66 |
==> PROPOSAL(just cleanup) |
| 67 |
|
| 68 |
- implicit/automatic writes/read to fs |
| 69 |
- explicit interfaces are better then implicit |
| 70 |
- same problems as in cdist 1.x to 2.x move! (environment!) |
| 71 |
- format on disk should not change/dictate code flow |
| 72 |
- degrade python to shell (nico++! steven--) |
| 73 |
- user should not care about python, ruby, .net or ASM implementation (steven++ nico++) |
| 74 |
|
| 75 |
? proposal 1: diverge emulator / core |
| 76 |
- emulator verifies input |
| 77 |
- emulator writes to fs |
| 78 |
- core reads/syncs from/to fs before passing control to user |
| 79 |
|
| 80 |
? proposal 2: emulator is dumb and passes data to core |
| 81 |
- core creates objects |
| 82 |
- no fs involved |
| 83 |
- core reads/syncs from/to fs before passing control to user |
| 84 |
- passing: |
| 85 |
- full objects via pickle |
| 86 |
- parameters only |
| 87 |
- how??? |
| 88 |
- unix socket? |
| 89 |
- not everywhere possible? |
| 90 |
- tcp / ip |
| 91 |
- not everywhere possible |
| 92 |
- chroot / local only |
| 93 |
- rfc 1149 |
| 94 |
- not everywhere possible |
| 95 |
- missing avian carriers |
| 96 |
- 0mq |
| 97 |
- not everywhere possible |
| 98 |
- not installed |
| 99 |
- shm (ipcs and friends) |
| 100 |
- not everywhere possible |
| 101 |
- no /dev/shm, different libraries? cleanups needed... |
| 102 |
- what speaks against FS? |
| 103 |
- emulator_input/.../ |
| 104 |
|
| 105 |
- nico: to fancy probably |
| 106 |
|
| 107 |
! boolean implementation |
| 108 |
==> BRANCH[feature_boolean_parameter] |
| 109 |
==> PERSON[Steven] |
| 110 |
- nico: |
| 111 |
- parameters/boolean: document |
| 112 |
- argparse changes (consider parameters/boolean) |
| 113 |
- create |
| 114 |
- can be implemented with changes in emulator |
| 115 |
- use store_true, del if false => never seen by core |
| 116 |
- INDEPENDENT INDEPENDT OF FS.PROPERTIES!!111111! |
| 117 |
|
| 118 |
- emulator: |
| 119 |
- how much integrated into core |
| 120 |
- also: using CdistObject???? |
| 121 |
- dependency on filesystem: good (nico) | bad (steven) |
| 122 |
|
| 123 |
- singleton / support without object_id |
| 124 |
- not discussed |
| 125 |
|
| 126 |
- __apt_ppa: |
| 127 |
==> BRANCH[bugfix_do_not_change_state_in_manifest] |
| 128 |
==> PERSON[Nico] |
| 129 |
|
| 130 |
- logging divergent between emulator / core |
| 131 |
- no problem (nico) |
| 132 |
- may be helpful (steven) |