Re: some questions about PostgreSQL in general - Mailing list pgsql-novice
From | Josh Berkus |
---|---|
Subject | Re: some questions about PostgreSQL in general |
Date | |
Msg-id | web-540210@davinci.ethosmedia.com Whole thread Raw |
In response to | Re: some questions about PostgreSQL in general ("Jules Alberts" <julesa@arbodienst-limburg.nl>) |
List | pgsql-novice |
Jules, > > Darn! Too bad you're not in my area. This is exactly the kind of > work > > my company does ... > > yes, that's too bad indeed. support is hard to find over here, we're > located in the southern part of the netherlands. Well, if you get desperate, we could always set up some kind of online-only support thing. Possibly in coordination with PostgreSQL Inc. I'm not sure how their experience with Netware/Firewalls rates. > yes, i know. i don't know much about firewalls, SSL, etc. (my network > background is mostly NDS, netware, IPX), the best decision is IMO to > consult an external company for the setup and have another one audit > for security. that gives me more time to focus on the central issues > like programming, db design etc. Yes. And you're more likely to find a network security company in the Netherlands than a PostgreSQL or PHP one. > well, i guess i'm lucky: we don't use windows apps for business > critical apps. the main app was in clipper 5.3b which is very stable. > the reason we dump it, is that it's language oriented, not db > oriented, > and it's no longer developed. And there's that pesky Y2K bug. :-) > OK, i have (and already had) that etched into my mind. we had a 2 > year > pre-build stadium in which we (management team + management) > evaluated > our company and what we expect from our systems. This sort of ties in with your question about table re-design. There are some limitations to altering existing tables ... for example, dropping a column requires a table re-build (adding or re-naming a column, and adding constraints, are permitted). However, in my experience, significant changes to the main data tables in a running system usually require redesign of a number of components, so the table re-build is less arduous than, say, checking all of your scripts for references to the dropped column. Overall, it's better to get the data design nailed down to the 90% level before you write any SQL. And this is true regardless of DB system. > after some lurking on the postgresql mailinglists there seem to be a > few items in postgresql which should be avoided, like the array > datatype and BLOBs. i guess it would be good to keep the design of > the > db as straightforward and elegant as possible and avoid more exotic > features... Yes, there's chronic problems throughout the industry with the "advanced" data types (MS SQL Server has much worse problems with BLOBs than Postgres). Partly it's that these types were not in the SQL 92 standard, and some (BLOBs) depend heavily on the host filesystem, so that implementation is RDBMS- and OS-specific. Some DB theorists, notably C.J. Date and F. Pascal, are opposed to the use of non-atomic data types in any form and their arguments have significant merit. > as for the development time, maybe it's a good idea to build a more > sophisticated front-end for the power users in Java, and a quick-and- > dirty HTML frontend with PHP. no offence meant, maybe in the end it > will be PHP only. Ah. I did not make myself clear. In the PHP-vs-Java comparison, you have two different layer structures: Web Server | PHP + HTML | Postgres Functions, Views and Triggers | Data tables or: Web Server | JSP Engine + HTML | J2EE or Java Middleware | Postgres Functions, Views, and Triggers (optional) | Data Tables The difference here, between PHP and Java, is that Java includes an extra, robust middleware layer capable of sophisticated, modular business logic and data management across several servers, if desired. Also, display/interface programming (JSP) is seperated from business logic (J2EE or java libraries) making maintenence easier. The PHP/Postgres Function structure, while quite economical for rapid development of small database applications, lacks this robustness and flexibility. It's a development cost vs. long-term value argument, and I've done projects on both sides of the argument. BTW, you can do the same things using Perl::DBI as you can with Java. I'd also love to hear from a Python fan, if we have any on the list. -Josh Berkus ______AGLIO DATABASE SOLUTIONS___________________________ Josh Berkus Complete information technology josh@agliodbs.com and data management solutions (415) 565-7293 for law firms, small businesses fax 621-2533 and non-profit organizations. San Francisco
pgsql-novice by date: