The people have spoken and the new framework is a dud. Outside of iOS development the new framework is only used sparingly when you have no other choice. There are numerous actors speaking up about this on both the forums and via their personal blogs. Recently Bob Keeney of BKeeney Software addressed some pain points of the new Xojo.Core.Date class on his blog. Xojo CEO Geoff Perlman makes an effort of defending the choices around the new Xojo framework but frankly nobody is buying it and Bob Keeney illustrates this perfectly in the comments.
Xojo.Core.Date is certainly flawed and it is not out of desire for code purity or technical merit. It is flawed because it was designed by an engineer to be consumed by engineers. However, Xojo is a product which was designed for prosumers and small valiant teams of software developers aiming to be more productive.
It shares a lineage with Visual Basic in getting out of your way and making your computer as a tool more accessible. The fact that it is so capable and can be used by computer scientists and engineers to build beautiful software products is a virtue few similar products share. It has for a long time walked a tight line between exposing or abstracting too much of the underlying system. It is rewarded for that effort by a very strong and passionate community and continued operations outlasting most other tools of its category.
Sway too far on the side of pure technical merit and code verbosity and you find yourself in a situation where hobbyists, prosumers, and long-time customers no longer recognize the product. Alternatively, if you abstract away all the details, fail to offer plugin SDK’s for all targets, and move away from Declares to an unknown “Interops” system you risk losing the professional users and barely-hanging-on third party ecosystem.
I keep repeating this idea in my posts that I do not understand why all the mystery that surrounds the Xojo product and its ongoing development. Why all the subterfuge about priorities and goals. Xojo will say “We want to support as many targets as reasonable.” Cool. We accept that. “We will support Android sometime in 2017.” We have no screenshots, no class hierarchy, no design goals, nothing to base any kind of opinion on what that product will look like. It is difficult to get excited or plan for the future when many attributes of Xojo iOS left us wanting more.
I know many who have attempted to build iOS apps in Xojo and decided it would be more profitable to learn Swift. Especially so when they were going to have to start over with a new standard library anyway. I am scared to think about what the Android platform might look like. Will it have the new framework? Or some unique variant of it with its own caveats and controls?
We seem to be moving towards a direction where Xojo is a cross platform IDE and compiler toolchain first and language second. All concerns about how the user actually writes their code and what that experience feels like and how productive they can be is secondary to shoving it into the IDE and calling it done. Maybe it was always that way but the framework shines a light on the self-imposed challenge they face.
Customers will put up with long term LLVM changes and delays. They will put up with an IDE that they prefer less than the previous version and a vague commitment to “fix it” in the future. They do fear it might actually get worse instead of better when the design process is completely secret. They can even deal with years of backlogged bug requests which are tagged as “Needs Review” or “Verified” with no fix in sight as they can often ‘Declare’ and work around it.
What concerns me is they may not put up with an entirely new language they do not even recognize. It is annoying to have to re-train for the new framework depending on what version of the IDE you are using and what platform you are targeting. Moving between projects is painful when I have to keep notes on which versions of Xojo has certain features working or broken. Feedback item #46943 with an essential performance issue on a CORE datatype to the language remains unfixed or even verified after a large release, a point release, and a beta release. How are bugs prioritized? Language data types functioning correctly would be #1 for me as without it you do not have a language.
If Xojo does not care that basic features such as data types are not working properly on all platforms then it gives us ZERO confidence in the new framework. As customers do decide to embrace the new framework, they discover features that are broken or missing on one platform or another. It is simply not tested very well and not used internally and only exists for the sake of iOS.
Somebody needs to take responsibility for the language and trash the new framework or commit to it. Especially considering there is no reason why you could not simply replace the Classic underlying implementations with the newer ones to give us the modern features with our existing code bases.
HTTPSocket could be an alias for HTTPSecureSocket which could essentially just be the new Xojo.Net.HTTPSocket behind the scenes. That is called abstraction and that is what customers pay Xojo for: A unified development environment and standard library across ALL platforms.
As it stands this is how many new programmers get introduced to HTTP sockets on the forums:
-----
New Customer: Hi! I like Xojo and I’d like to query a web service and get some results. I tried using the HTTPSocket but it is not working.
Old Customer: Hey great to have you. Well you should use HTTPSecureSocket because it supports SSL and most web services require SSL these days.
New Customer: Oh great I had no idea that even existed. Why the distinction? Isn’t SSL ubiquitous?
Old Customer: Yeah Xojo did not change HTTPSocket in the off chance that a customer desires HTTP capabilities without SSL functionality. We have not found that customer yet but alas we are still looking…
New Customer: Okay thanks. I tried it but it does not work. Something about authentication error. The web service says it uses digest authentication because basic authentication is less secure. I tried to use the ‘AuthenticationRequired’ event but no go. Tech support said I also need to use HTTP 1.1.
Old Customer: Ah yeah so for that you want to use the new Xojo.Net.HTTPSocket class (there is no Xojo.Net.HTTPSecureSocket) because it supports digest authentication, SSL and HTTP 1.1.
New Customer: Ah wow makes sense. Okay so I have a problem and that is I can’t seem to get my data back. Where does the response string get returned?
Old Customer: The new class is asynchronous which means you need to entirely change your thinking and handle the PageReceived event. Best ways to do that is drag an instance of Xojo.Net.HTTPSocket to your window (actually drag an object and rename the super because the new classes are not in the library…) OR subclass Xojo.Net.HTTPSocket
New Customer: Wow ok thanks.
-----
The IDE outside of iOS makes no effort to promote the new framework. You would not even know it existed unless you were working around an issue and someone suggests a new framework class. If you are not working around the classic framework then you are actively discouraged from using it at all. Many users on the forum repeatedly say to avoid the new framework and even Xojo engineers remind us that Classic is not going away anytime soon so no need to invest in it today.
May I ask what is the point? Why does Geoff and company work so hard to convince us that the new framework is technically better and that we will “get used to it”? They make no effort to commit to it or even verify it is working correctly on all targets. What is the endgame?
If you are going to commit to it then lets formally deprecate or remove the Classic framework and force users to move forward or stay behind. It is a simple choice and the upcoming capabilities of the platform should sell itself. To make that process of converting easier add in some helper methods and properties that make the new framework feel like the old one.
For example you could add a TotalSeconds property to Xojo.Core.Date and abstract away the details of the Xojo.Core.DateInterval class to better support existing code. You could have a SQLDateTime() method on Xojo.Core.Date so we don’t have to run to the documentation to figure out how to get a Date into a Database.
You might be right that it can be done with extension methods but if all your customers are using those methods now then why not just add in a Classic extensions module to every new project (or provide a checkbox when starting a new project – “Support Classic API?”). It then automatically adds in some Classic API support to the new framework via extension module to make converting easier.
Or if you are not going to commit to the new framework and make it easier to use then please call it a dud and get rid of it. Many customers rely on Xojo to abstract away the details. They do not want to be introduced to three different ways to initiate an HTTP socket with various pros and cons. Especially when in that case the old sockets are simply obsolete and should not be used.
Behind the scenes you can make the Classic classes use the new framework equivalents. To the end user the API never changed and the value of the platform continues to improve as you evolve the standard library. Now that is a subscription worth paying for.
However, if you decide that the computer scientist and code aficionado is your target market then be prepared for the new framework to die a slow death and Xojo’s future with it.
Introducing the new framework was Xojo’s very own Windows 8 moment. Now it’s time to bring it all together in a way that reinforces the Xojo value proposition:
One development environment. One language. One standard library. Many targets.