Nguyengia, I think Luis' objection to using an SQL database is regarding not using simpler methods like an hashmap implemented in C or via shell scripts.
What he's unable to understand is the complexity involved in figuring out the dependencies associated with a particular package and therefore their management. This complexity is easy to handle using the declarative language; SQL, which is based on relational algebra. An alternative could be to use reasoning techniques like SAT solvers or inference-engines. Cheers. On Sunday, August 01, 2021 03:50 PM IST, "hung.nguyengia via illumos-discuss" <[email protected]> wrote: > What's wrong with SQL my friend? I subscribed to this mailing lists only to > read but your statement about SQL is totally wrong so I wrote my answer. > Check FreeBSD's pkg package manager (same name but different), it uses a > sqlite database and I know many package managers on Linux also use sqlite. > Sqlite is indeed designed to be used with such scenario, being an embedded > database. Hope it helps. > > https://github.com/freebsd/pkg > > Sent with ProtonMail Secure Email. > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ > > On Sunday, August 1st, 2021 at 3:52 PM, <[email protected]> > wrote: > > > G'day illumos people. > > > > I was studying the SRV4 package management-- i want to re-implement it, but > > i don't > > think talking about this matters now-- using Heirloom's as reference and > > i've found that > > it uses SQL in the libpkgdb-- apparently for register installed packages. > > Then, i've checked the illumos' one and saw that it doesn't use SQL and, > > instead, > > it uses a plain-text file that is parsed by a function in pkgserv.c > > (possibly I'm wrong, > > as my C skills are really basic). > > > > What exactly that SQL databases in "original" pkgtools used to store > > (besides registering packages installed)? I haven't read all the code yet, > > so if you could tell me it would help a lot. > > > > Why did people at the Solaris development group used SQL rather than some > > simple > > database implemented from scratch or a simple hashmap? And why did you guys > > replaced it-- as illumos' goals aren't exactly ultra-minimalism a la cat-v? > > > > I'm afraid if this question sounds "dumb" or something, but i really wanted > > to > > talk with people that actually read all the code and maintained it instead > > of > > taking a mistaken approximation based on my guesses reading some code and > > docs about it. > > > > Sorry for any "bad engurisshu", although reading/writing in english for > > more than five years, it still sucks. > > > > Sincerely, Luiz. > > > > illumos / illumos-discuss / see discussions + participants + delivery > > options Permalink ------------------------------------------ illumos: illumos-discuss Permalink: https://illumos.topicbox.com/groups/discuss/T5f71878f66d3ba97-M968eccff272c19e68c92d7ad Delivery options: https://illumos.topicbox.com/groups/discuss/subscription
