Web Based Multi-User Interface: Not a new subject. (1 Viewer)

Thales750

Formerly Jsanders
Local time
Today, 10:06
Joined
Dec 20, 2007
Messages
2,150
500 thousand lines of code later. For 2 decades it has been foretold. Now we have seen the asteroid.
It has arrived.
The end of manual code writing, and the end of operating system based programming.
In this review and rehashing of an old subject, I would like to have a discussion on Web based databases with Browser interfaces. All done by AIs, methods, pitfalls, advantages, all of it.

What are your thoughts? Don't assume ask for clarification as necessary.

How would you use AIs to build a full scale program? Say, something like an ERP for a midsized company? Something more than asking Chat GPT to write a procedure or a function.
 

Pat Hartman

Super Moderator
Staff member
Local time
Today, 10:06
Joined
Feb 19, 2002
Messages
43,491
I would like to have a discussion on Web based databases with Browser interfaces. All done by AIs, methods, pitfalls, advantages, all of it.
At some point the programmers who wrote the AI will have gotten smart enough to make a flawless AI but since the AI has been created by flawed programmers, we are not there yet. Just like we are not even close to having the necessary infrastructure to have an Electric Grid powerful and uninterruptible enough to heat our homes in cold climates, heat our showers, light our homes, cook our food, and power EVs.

People are obsessed with browser based apps because of their ease of distribution and their ability to support remote users, not for any other reasons. I have never seen a browser based app that can hold a candle to one built using client/server technologies. Most actual browser based apps that I interact with as a user, suck. Express Scripts is just about the worst purpose built application I have ever been forced to use. It's property settings are illogical ( I can hide active prescriptions but not old ones with no refills left), it sends me stupid, useless emails with no specific information (for my safety) that I can say, "oh, my xx med has been refilled and will be sent next week". Instead, two days after a 3-month prescription is shipped, I get an email telling me that my refill (for the next fill) is in process. Why?
How would you use AIs to build a full scale program?
Have you ever seen a spec that was detailed enough for you to build a complete application? If we can't even create a spec to that level, how can we tell an AI what is required? There are a number of business applications that are fairly well defined but even those have an enormous number of variables that need to be defined to set all the properties correctly.

Baby steps. Let the AI help you to optimize code. Let it build common code libraries that you can help it to put together. At this point - never trust an AI to write code that you do not completely understand. They make things up. All experienced developers have dozens of templates and patterns and complete subsystems that they incorporate in new apps. This is why we can be so much more efficient than those who have to reinvent the wheel for every new application.

When I was managing development teams, my staff ranged from "could not be trusted to fetch coffee" to "we could shake hands and they would understand their task". At the moment, AI is closer to can't fetch coffee than to being able to produce something even rudimentary unless the business process is one that is so well defined that they teach college courses for it.

A test - there is a very specific basic pattern for creating a questionnaire. Without letting on the implementation details, see if you can get an AI to create a correct schema. Start the way a user with no knowledge of relational tables would. I have 120 questions.... A couple of years ago one of my clients wanted to upgrade their questionnaire (which I hadn't built, someone else had) so I gave them what I thought was an outrageous price to build it with Access (the client was a real PITA so I knew there would be multiple redesigns) and a 4 month time because I was planning on bringing on a second person to help. The App would be built with Access and run off-line. The visiting nurse would enter the answers on her laptop and then upload the answers when she came into the office so no additional hardware or software would be required. They didn't even want to discuss it with me. They went with a web app. The bid was $3 MILLION dollars (boy was I stupid). It required them to buy iPads. It required them to buy hotspots for their phones for a monthly charge of $50 because the nurse couldn't go into the home of a poor person and expect to have a live internet connection so she needed to bring her own. Why? Was any action going to be taken immediately? No, the answers needed to be reviewed and that always takes some time and only then could a care plan be created. Real time was just a waste of money. It required $500,000 for annual support. and they wanted 2.5 years to build it. It took over 4 years and cost $5 million. There were 80 question tables plus 80 answer tables + 80 separate web pages with hard coded questions because the developers did not know the questionnaire pattern.
 
Last edited:

Thales750

Formerly Jsanders
Local time
Today, 10:06
Joined
Dec 20, 2007
Messages
2,150
People are obsessed with browser based apps because of their ease of distribution and their ability to support remote users, not for any other reasons.
I'm agreeing with most of what you just said, except:
The Users prefer the interface.
This system MUST be mobile.
Access has actually fallen behind on User Interface widgets. It's probably a result of a hundred million more programmers in that arena. I don't know about you, but other than a handful of small systems for local only companies, we spent the last few years working on Chemical Plant UPGRADES and UIs because they have SQL in all of their disparate systems.

Our other large system, the one we will be converting to Browser based is for Design, Engineering, and Construction, I currently have some pretty good mobile methods using Outlook local, Outlook.com, and One Drive. It works ok, but it is not scalable.


I don't need an AI to write an Access program or an SQL database. In both cases though, using the AI for Procedures, Functions and some query stuff works great. It's mostly Functions, Chat GPT really gets functions.

No Form modules, that would be more effort than it's worth. On this part I totally agree with you. I can't just walk in and ask Majel Barrett to write a program, now can I?

But I want to learn how far if can be taken, that' why I started this thread.
 

GPGeorge

George Hepworth
Local time
Today, 07:06
Joined
Nov 25, 2004
Messages
1,994
I'm agreeing with most of what you just said, except:
The Users prefer the interface.
This system MUST be mobile.
Access has actually fallen behind on User Interface widgets. It's probably a result of a hundred million more programmers in that arena. I don't know about you, but other than a handful of small systems for local only companies, we spent the last few years working on Chemical Plant UPGRADES and UIs because they have SQL in all of their disparate systems.

Our other large system, the one we will be converting to Browser based is for Design, Engineering, and Construction, I currently have some pretty good mobile methods using Outlook local, Outlook.com, and One Drive. It works ok, but it is not scalable.


I don't need an AI to write an Access program or an SQL database. In both cases though, using the AI for Procedures, Functions and some query stuff works great. It's mostly Functions, Chat GPT really gets functions.

No Form modules, that would be more effort than it's worth. On this part I totally agree with you. I can't just walk in and ask Majel Barrett to write a program, now can I?

But I want to learn how far if can be taken, that' why I started this thread.
You're on the cutting edge, here. I might even say the bleeding edge. Let us know how you get on, though.
 

Pat Hartman

Super Moderator
Staff member
Local time
Today, 10:06
Joined
Feb 19, 2002
Messages
43,491
But I want to learn how far if can be taken, that' why I started this thread.
@Uncle Gizmo Is obsessed with it and regularly posts in a watercooler thread with his adventures. We've had some good laughs at the AI solutions though. It frequently just makes stuff up out of whole cloth. That is why you MUST be competent enough to either recognize that the solution will not work or test it well enough to judge.

I would have a whole lot more faith in AI and its progression if it is taught to say "I don't know". Very simple --- "I don't have enough data to come up with something I would trust but I think this solution will help you to get where you're going". "Ask me some more questions and I'll try to refine my solution". "These are some links that may guide you". Whatever. It's like politicians and the media. When you KNOW for a fact that they are lying to you because you can see the empirical evidence, it is very hard to trust anything they say. At least an AI doesn't have an ego to feed so it should be able to learn to recognize what it "knows", what it "doesn't know", and what "it thinks might work".

I try to distinguish between 1 and 3 when I answer questions. For 2, I tend to stay out.
 

Thales750

Formerly Jsanders
Local time
Today, 10:06
Joined
Dec 20, 2007
Messages
2,150
You're on the cutting edge, here. I might even say the bleeding edge. Let us know how you get on, though.
LOL. My little company designs, engineers and does the building Science for Coastal Homes that have to stand up to 150 mph winds. My whole life is at the tip of the spear, where bloodletting is common.
 

Edgar_

Active member
Local time
Today, 09:06
Joined
Jul 8, 2023
Messages
438
How would you use AIs to build a full scale program? Say, something like an ERP for a midsized company? Something more than asking Chat GPT to write a procedure or a function. (adsbygoogle = window.adsbygoogle || []).push({});
For web development, choose a stack and a database. So ask the GPT this:
1. I come from ms access, show me some frontend/backend stacks to start with web development
Regenerate the answer multiple times. You will notice a pattern that will help you see what is the most used out there.

Then, to choose a stack, ask the GPT something like:
2. Show me a minimal example of a frontend file and a backend file for each stack and explain what I can expect from the example. Make sure to include how the html and css files will be generated
Again, regenerate the answer multiple times, it will output a lot of stuff, so it might skip some stacks, make sure it does not skip stacks. That will let you see very quickly what you may want to dive into, for instance, I hate Angular, but love React.

Once you choose one that looks nice to you, ask something like:
3. how should I arrange my files for ChosenStackName?
Regenerate, again, you will notice patterns. You will see all kinds of folders and file names, just as your questions about them.

Once you're done with the file structures, ask the GPT:
4. So which folder or file from that structure should I start with?
Yes, regenerate as many times as you can, in fact, you can make suggestions of file structures you might find easier for yourself.

The GPT now knows where you're going, you can ask it something like:
5. Show me the steps to show something on the browser using that stack right now
You know the deal, regen until you notice the patterns. Now you know what to expect, so go watch some tutorial of someone setting up an app with that stack. And make adaptations.

Make it work. Copy the tutorial, whatever, just make it work. Then, ask the GPT:
6. How can I adapt this file for an ERP app? and paste the snippet of code from the tutorial.
Make your adaptations. For example, if in the tutorial, they were using MongoDB and you want to use MySQL, ask the GPT for those adaptations.

This is just an example. When I started there was no GPT, but after watching the outputs to these questions I assure you it will give you very decent results. I'm in fact thrilled about some of the responses and I'm already questioning my beliefs. This is absolutely normal in web development, there's a lot of hype everywhere, but if you focus on one stack and you stay there, it will give you results. Problems arise when you're always looking for the newest, shiniest framework. The antidote is to focus on mature, proven frameworks instead of new ones. Old is not always bad, if it's old and relevant today, it's because it works.

Regarding the topic of this thread, the GPT can help you tons here. But it only has a certain "scope" to work with, you must be the captain in charge of the entire project, you can't upload all your workspace and expect it to understand it all for you (not yet, anyway, give it a few months, hint: it won't be free). I pay for copilot, but even though it's supposed to read my workspace, it can't quite do the thing yet, it's still very good suggesting autocompletion, it's good file scoped, but not workspace scoped. So, again, make something work first, then ask the GPT to make adaptations to the thing that already works, file by file. For that, it's very good.
 
Last edited:

Uncle Gizmo

Nifty Access Guy
Staff member
Local time
Today, 15:06
Joined
Jul 9, 2003
Messages
16,365
As Pat mentioned, I have had some interesting adventures with chatty...

I'll tell you about it below, but first an answer to your question (of sorts)...

I believe the essence of your question is about putting a database on the web. This is my answer to that question.

I've posted this on AWF before:-

One of my rainy day jobs is to put a database on the internet using PHP and MySQL. Haven't done it yet, and I probably never will! Still, it's an ambition of sorts.

I did an online course which I thought was fantastic! I highly recommend:-

Robert Tucker of Infinite Skills

I found this PHP MySQL course exceptionally clear and easy to follow. I particularly like Robert Tucker of Infinite Skills, his voice, I could listen to him all day!

The course is currently free on Udemy, and I have seen it in other places on the internet for free as well. I also found the content, the sample files somewhere for free, not sure if I could find them again…

Beginner PHP and MySQL Tutorial

You couldn't go far wrong in taking this free course, to my mind it's exceptional.

I made this note about a “Framework”
(Use a framework)

Regarding chatty, I was discussing the various options of how to open a form and pass information between two forms. A pet topic of mine!

Chatty told me to set the custom property of the form I want to open and then open the form and I would have access to the information in the custom property. I knew immediately that this was wrong!

I told it so, and it gave me one of its usual groveling apologies.

It talks too much! Even when you try and tell it to shut up a bit.

I wonder if it's female?
 

AccessBlaster

Registered User.
Local time
Today, 07:06
Joined
May 22, 2010
Messages
5,999
@Edgar_ has made clear, you have to keep redefining your questions. You see that technique everywhere even here. Seldom does an OP or expert understand the question immediately. It usually takes a lot of prompting for both sides to understand the situation.
 

Thales750

Formerly Jsanders
Local time
Today, 10:06
Joined
Dec 20, 2007
Messages
2,150
I'm a little old for all this learning. But what else is a guy supposed to do?

Thanks All.
 

Users who are viewing this thread

Top Bottom