Re: transactional file persistence?

Tommy Halsbrekk wrote:

Lew wrote:
Tommy Halsbrekk wrote:

indexes and miscellaneous resource management components. Any of which
increase the complexity and can fail for a number of reasons.

Red herring. Well designed DBMSes have that covered.

So by your definitions, databases do not ever fail?

These days, data corruption in a RDBMS only occurs due to user error,
like messing with system tables directly, making a file based backup
while it's still running etc.

Its a use case with requirements for an entirely different use than a
db. Just because it requires transaction support does not mean the only
solution is a db.

No, but it does suggest itself, since your requirements sound very much
like one. Why not look into it some more, but you seem to be hostile
towards the idea.

- one solution is easier to fix than the other.
- one solution requires less expert knowledge to fix than the other.
- one solution has less components that can fail than the other

and the file system wins in all of these cases for my requirement.
So the conclusion is if one does not need a structured data store, then
dont use a db.

Some RDBMS, like Berkeley, are quite "simple" IMO. Subversion uses it,
and it does not seem to have much overhead at all.

But, if you say this, I don't get why you asked "for ideas". You have
made up your mind, so run with it. We won't stop you. ;)

It is fairly common for a db to have some sort of problem, which
prohibits completion of the operation. It is not very common for a
filesystem to experience the same level or complexity of problems. The

You are wrong. Filesystems are notoriously unreliable for transactional
operations. DBMSes, the good ones, have all kinds of safeguards and
features to support data integrity and durability. Advantage: DBMS.

Are there any reason I cant make use of any of these safeguards in the
solution I make, without using a db? or is that prohibited for some reason?

Not at all. But you did ask for ideas.

I know that what you are trying to say is that I should use off the
shelf components, because its tested and works.

You might want to re-visit your engineering text books, specifically the
chapters about adapting your solution to the actual requirements. Using
a sledgehammer to kill a bee is not the solution to the problem I am
trying to solve.

It all depends on the situation of the project. Sometimes, it
compensated adapting the sledgehammer instead of constructing one
yourself. Textbooks are often not very close to real world scenarios.
YMMV.

And don't kill bees. We need them for food (polonization). <g>

.

Relevant Pages

  • Re: Any Substitute for READONLY in F77?
    ... about any I/O operation can fail. ... reason in Fortran why this one should. ... My reading of the OP's request was that he appeciated the protection of ... net worth using. ...
    (comp.lang.fortran)
  • Re: Choice of Language for FPGA programming
    ... well-managed and easy to reason about. ... Nonblocking assignments get you within ... been whining for years about their dislike of the complexity ... DOULOS - Developing Design Know-how ...
    (comp.arch.fpga)
  • Re: PTHREAD_MUTEX_INITIALIZER
    ... reason internal to the implementation, so you need to check the return ... anything sane if locking a mutex fails for no application-level ... Locking that mutex might not fail, ...
    (comp.programming.threads)
  • Re: FWD: Latest update on Pat "Nine Toes" Flannery
    ... No women over ten years older than you are for the same reason. ... No women whose ex-boyfriends were crazy or criminals, that should be a real good clue that there is something odd going on upstairs with them. ... as that adds a great deal of complexity to any relationship.* ... No women whose finances are in a state of constant chaos, as that indicates that your finances will also end up that way shortly. ...
    (sci.space.history)
  • Re: I want six signatures
    ... I will not fail. ... I have tried to fight the mood and the urge to smoke over the last three ... In the past smoking had helped to get me out of the ... Is that enough reason for you ...
    (alt.support.stop-smoking)