Formal policy for project to be a software partner

imagej
slim-curve
openspim
policy
scikit-image

#29

Dear Forum
(Thank you for the input Herbie)

It is certainly possible for this to happen. It depends on where the question is posted for example (on the company forum or on another forum). It will also be influenced by whether the user posting is a known user with a valid tech / app support - those naturally get priority. There are several other factors.
My original comment was specific about the company I currently work with - i.e. the one I can control to ensure we would follow up on tech/app support requests on this forum.

When visiting microscopy core facilities and individuals labs around the world I see they typically use several pieces of software from free and open source all the way to commercial and closed source to complete a pipeline. Many tools (free, commercial, open and closed) aim to provide users a complete workflow. The reality is users select a few tools and string them together to complete their work (often including one ore more commercial tools along with FIJI apps and other great tools).

As I mentioned above, this forum was created by and for open source tools. At the start of this thread / discussion it looked like the forum was considering to include other tools. Also, on twitter (where I originally saw the news) it looked very much like it was an effort to bring all tools under one roof (not just the the open source ones). see for example: https://twitter.com/cmci_/status/1025262569710186496 or more generally, https://twitter.com/search?vertical=default&q=forum.image.sc&src=typd

The more vocal users on this thread have made it clear that the forum should be for open source tools only (commercial or not is another question). Naturally, I respect those opinions - everyone in the community has a voice. I will continue to do my part to make the wider bio image analysis community a stronger / better / more impact-full one.

All the best
Luciano


#30

Hi all,

I really hope that this discussion leads to balanced opinions on both sides.

I really do not think that any company sees this forum as a replacement for their own support etc. At least that is true for the one I am working for.
But as already explained most people use free and commercial tools, closed and open source.
So if someone has a question and somebody suggests a solution that involves a commercial solution, why is this a problem? And even if it would be closed source it is still up to the user to decide which software to use.
And the fact that companies make money is not something bad, because money for research and grants must finally come from somewhere, isn’t?

So please set up clear rules what companies can post here. And certainly I encourage discussions about used algorithms etc. But that does not automatically mean that every company should publish their complete source code. Or is this expected?

Sebi (from Zeiss)


#31

To reiterate @thewtex’s lost point: we should certainly distinguish between commercial/for-profit, and closed source. ITK/Kitware is a great example of a for-profit company benefiting from and giving back to the open source community. In contrast, Zeiss (sorry to pick, Sebi) recently released Intellisis, a blatant clone of Ilastik, while making zero contributions back (as far as I can tell, would be delighted to be corrected).

I’m sure many in this forum have been frustrated with the difficulty of getting funding for open source development, while at the same time the very departments that employ them spend tens or even hundreds of thousands per year on licenses for closed source software.

In my opinion, this forum should be used to promote open tools, and thus lift all such tools. (And promote their interoperability.) This is not a strong opinion, and I’m happy to see the alternative side of everyone getting along. It absolutely doesn’t have to be a political decision or statement. But I would prefer if it were.


#32

Hi,

you don not have be sorry to pick the example of Intellesis from Zeiss, since I like to have an open discussion with people that disagree with my view.

But to be honest I do not think it is fair to call this software module a blatant clone, without knowing all the details of that tool.

  • it uses Python 3, Dask, Scikt-learn, Tensorflow and … which is all open source
  • in the process of developing it we reported a few bugs to dask that got fixed. Not much, but not nothing …
  • we make no secret about what we do in terms of algorithms
  • it uses a completely different way to handle multi-dimensional datasets compared to other tools
  • this software is all about integration in the existing measurment framework which others tools obviously cannot offer
  • And there a many other differences in the details

Just because it does pixel classification it is not a clone of ilastik. This is way to simplistic. Is every software that does pixel classification a clone of ilastik or in case of counting cells or stitching a clone of Fiji?

If that would be true many tools in the open source and free software world are also just clones of something developed by tax-payers money that could have been spend on something else. But I really like the diversity of tools and the option to choose what works best for me.

Seriously there are tons of tools in all areas that do similar things using partly the same ideas? Are all those useless and juts “clones” of someone favorite solution? Certainly not.

All those things were requsted by our customers to solve their issues with existing tools for their applications and workflows. Such tools are not developed to create a “clone” of an already existing solution but because their is a real need for that.
Moreover Zeiss does support especially Bio-Formats (it is a paid add-on of ZEN where licences fees have to be paid to the creators of Bio-Formats) and is involed in many research projects. Of course at some point as a company it has to earn money.

So what woukld be your idea of a real contribution that would allow a company to develop a tools that use a basic ideas that is already implemented somewhere?


#33

Hi @jni and @sebi06

I liked both your replies because I think these are the types of conversations that I think should be on this forum. Having an open conversation about the strengths and weaknesses of tools and development philosophies is always good.

As a side question did Zeiss ever consider developing open source tools?? Did anybody look closely at the economics of this?? How much revenue does the software generate as compared to hardware? Perhaps a difficult question to answer…

@thewtex I’m curious for your input on this as well. Are you aware of any analysis of open source vs. closed source business models for hardware companies?? I realize that is a pretty general question, but still curious.


#34

Yes, there are plenty of open source software business models.

  • A hardware company selling sophisticated equipment can sell the hardware and provide supporting software open source. It is even possible to have open source hardware. Here is a nice, recent review of Emerging Business Models for Open Source Hardware.
  • Infrastructure services can be provided for open source software, providing the necessary hardware, software maintenance, and software integration. As imaging moves to the web, there are more opportunities here.
  • Companies, e.g. Red Hat, can sell support subscriptions for open source software.
  • Collaborative consulting services around open source software expertise is possible, e.g. Kitware.
  • Education and training services around open source software can be provided for a fee, e.g. universities, on-site training, online courses.

Closed source licensing models are often more profitable in the short term; in the long term, the cost, quality, flexibility, and capability of open source software wins out. We have seen this in microscopy with Fiji, which many scientists now simply prefer over closed source alternatives. We have seen this with Microsoft, which was extremely profitable in the 20th century with a closed source business model and actively hostile towards open source software, but it the 21st century it has been forced into an open source business model, i.e. Azure cloud computing services.

But, I hope this forum is agnostic to making money.

I would like to see a focus on imaging, science, and computing. Closed source software is sometimes required for current imaging workflows. However, closed source software does not allow inspection of the algorithms involved, extending the software for new research purposes, and sharing that software with other researchers so they can reproduce and understand the work. Open source software is required to reproduce and understand an experiment – it is necessary to do actual science.

To help strengthen the clarity of our purpose, we could

Enable open science and reproducible research.

and

Foster not only scientific independent thinking, but just as importantly, independent learning. We want to not only teach people how to fish, but teach them how to learn.

as stated in the ImageJ Wiki and previously emphasized by @imagejan.


#35

Those are good arguments, but in reality life is not that simple. In microscopy i really do not see many of those open-source buisiness models that are commercially sucessful (yet?). That might change but currently at least one reason is that many research institutions are not willing or are not allowd to pay for service and infrastructure.

It rather encountered many times the issue that the idea of “free” software but paid service, maintainence and infrastructure was rejecteted from the resarch institutions, because the way grants etc mostly work today only allows for one-time payments.

Regarding the idea of providing infrastructure for open source computing it might be worth checking out www.apeer.com (by ZEISS), which is a Docker and Kubernetes based image and data analysis platform that allows everybody to use their own code and tools.

And I have to make an additional remark here. We all love microscopy and image analysis (at least I do) and for years I had the assumption that most people using those techniques for their science think the same way, but reality is differnent. What many people want are just tools that get the job done. They do not have the time or the skills or the will to do a deep dive inside used techniques and algorithms on their open source tools or commercial tools.

As an example, how many people using a threshold or stitching (not speaking of things like DCV etc.) actually did check the used implementation inside their favourite open-source tool? In theorry everybody could do it but in reality most people just trust those tools and use it the same way they would use a commercial tool, where they cannot check out the source-code easily. People here on the forum are in average much more on the expert side of life than “average users”. For most of them the idea of checking the (open source) code is something which sounds cool, but will be rather unrealistic for many (good) reasons.

So did I understand it correctly that the main issue is not commercial or free, but closed-source or open-source, because every scientist must be able to “check” the used software in case of … ?


#36

So did I understand it correctly that the main issue is not commercial or free, but closed-source or open-source, because every scientist must be able to “check” the used software in case of … ?

It depends on whom you ask…

I wrote (for good reasons):

No commercial products on this Forum!

Please read the whole thread!

Regards

Herbie


#37

Hi,

i did read the complete thread, but it leaves many open questions. Much more than it answers, because some arguments ignore real life facts and over-simplify things.


#38

i did read the complete thread […]

But you seem to have not considered my plea.

[…] some arguments ignore real life facts and over-simplify things.

Agreed!

Herbie


#39

Maintenance and infrastructure contracts are difficult to obtain from research institutions.

Practical approaches to work with research institutions:

  • Charge the actual cost of the hardware as part of the purchase and support contracts. I.e. do not require hidden software license fees to make the hardware useful.
  • Collaborate with grant PIs during the proposal period. Carving out funds later is often not feasible.

This looks cool.

The reality is that many people find the closed source tools limited and broken. And there is nothing they can do about it. Furthermore, when they want to re-analyze their data a year or two down the line, the data, which they may have spent have a decade of hard work to produce, is effectively lost because licensing funds are no longer available or the vendor simply decided to stop maintaining the closed source software. These are the real-life facts.

It happens often. Recently, in this thread, an attempt was made to compare Intellesis to Ilastik. It is not possible to have a solid understanding without the Intellesis source code. How, exactly, do they differ? How should we consider results created by the different algorithms and implementations?


#40

Especially when it comes to a tool like Intellesis feel free to talk to me about details since I am in charge of this software.

So far just the statement was made that this is just cloning ikastik, which is not really the case. Yes it uses partly the same python modules but calling this a clone is a bit to harsh.

And this tool is never advertised claiming it has magic algorithms. It explicitly uses proven and estabshed open source libraries and does focus on UI, data handling and integration into ZEN measurement pipelines.

So it is a hybrid between both worlds. So let me know if you expect detailed info about such a tool here in this forum. Of yes I will do my best despite working for commercial company.

If this is wanted because its is commercial, this is also fine with me.


#41

Hi, the point is not how many people do actually check the code, but that the process is transparent and that it can be done (and so it happens that checking is very often done). Many pairs of eyes see more than one pair. That is how many bugs are found, code is questioned and bugs are fixed, in my experience, way faster than commercial software that I have used.
One could use the same example and wonder how many bugs are there in software that nobody can tell what it does. Or whether bugs are too complicated or costly to fix, being left in a waiting list without notification.


#42

Hi @gabriel

I agree with you completely that some commercial companies let bugs and issues lie waiting because it’s not profitable to fix them. However that’s all the more reason to welcome commercial companies.

While some bugs are found by examining the code, many (perhaps more) bugs are found by very carefully examining and discussing results and output images. If a user of a commercial product has a concern, they can post the images on this forum and get feedback from experts in image processing. If there is an issue, people can suggest open source approaches that may handle the problem. In turn the commercial company will either fix the problem quickly, or get bad publicity.

My suspicion is that the only commercial companies who will actively post here, are those who already have very good support, scientifically rigorous testing, and responsive bug fixing. In turn these are probably the companies that would be most receptive to some of the business models @thewtex has suggested. So it seems to me allowing such companies to post will lead to a positive cross-pollination of ideas and knowledge.


#43

Let’s hope that the cross-pollination doesn’t turn out being a cross-pollution!

Brian your statements are full of hope and subjunctives.

We will see pollution by a great number of posts that are of the type:
How can I perform abc with xyz?

Where abc is an image processing problem and xyz is a commercial software that only a small number of the present Forum members sufficiently knows.

A consequence is that the Forum becomes clutterred and perhaps later may become increasingly visited by competent users (and makers) of commercial software.

Do we really want an extension to this clientel? I definitely don’t want to see it.

And finally, as Curtis has pointed out: Presently, companies that run their own help and support Forums are not considered being software partners of this Forum.

Regards

Herbie


#44

There is a lot to answer to here but I don’t want to nitpick so I’ll answer what I think are the two main questions from @sebi06:

So what woukld be your idea of a real contribution that would allow a company to develop a tools that use a basic ideas that is already implemented somewhere?

Well, you mentioned that you use Dask. Bug reports are useful, but since you are (presumably) making a decent profit out of Dask, perhaps you should devote resources to fixing these bugs yourselves.

Perhaps some percentage of the company profits should be donated to NumFOCUS or the Python Software Foundation so that the open source tools that you depend on can be better supported.

But, most of all, it would be nice if Zeiss open sourced Python packages for reading and writing .czi files, as just one small example. Or tools to integrate ZEN and Fiji/Python workflows. (Ditto, of course, for Leica, FEI, etc.)

So did I understand it correctly that the main issue is not commercial or free, but closed-source or open-source, because every scientist must be able to “check” the used software in case of … ?

Not every scientist will do so, this is true. But expert helpers will very often do this to understand what a particular bit of software is doing. The details matter a lot when trying to improve performance on a particular dataset. In other words, closing the source code is, at the detailed level, indeed “making a secret” of the algorithms you use.

I’m not ascribing maliciousness to this, just saying that the outcome is a real frustration for expert users, who often play in the 10% of cases where predefined workflows break in subtle ways, not in the 90% of “easy” cases.

As to business models, I am of course not a business expert. I can only say that when my institute comes to buy a new instrument, if a company charged a bit more for hardware but then did all of their software as open source tools that work nicely with the existing ecosystem, they would have a very persistent advocate in me. I imagine that I am not alone in this view.


#45

Thx for your answers which i really appreciate.

We already haven a beta version if a python wrapper of our open-source libCZI, which can be hopefullz released soon.

Reporting a bug is one thing but being able (time and skillwise) to fix it is something different despite the fact somthing is open-source.

And increasing the hardware price by Holding the costs for the software is tricky. How should that work for software only? Most big software platforms a microcope vendors are much, much more than just hardware control.

And i want to point out again, that at least i never had the idea to use this forum for commercial support. I just was curious to find out what is allowed in cases of questions posted here that have to do with commercial tools or etc.


#46

And i want to point out again, that at least i never had the idea to use this forum for commercial support.

What your ideas are and what actually will happen are completely different affairs.
If a user struggles s/he will search for help wherever possible and a Forum is independent of working hours, time zones, availability of a company’s experts, etc.

I must admit that now, after some days of fused Forums, I still don’t understand what CellProfiler has in common with ImageJ, except that ImageJ can be used for similar purposes. In fact, both systems are extremely different in their architectures and goals and I can’t see a direct software link between both. While ImageJ is quite universal, concerning the scope of applications, CellProfiler, as the name suggests, is more specialized. Both systems have their clientel and are justified but I would much prefer if they had their own Forums.

There must be a reason for getting bigger Forums that I presently don’t fully understand from a technical point of view …

Regards

Herbie


#47

Thanks to everyone for the vibrant discussion so far. I am really excited that we are having this conversation!

:sparkling_heart:

Speaking as one of the founders of this forum: we (the teams at LOCI and Broad) did discuss how commercial software would fit into the combined forum. Our intent is that discussion of commercial software, both open source and closed source, is welcome here.

Based on this conversation, I have expanded the relevant FAQ entry as follows:

And I have revised the FAQ entry “What is the forum’s mission?” as follows:

It is important to distinguish between embracing discussions of closed-source software vs. allowing closed-source community partners. My concerns regarding the latter have been expressed eloquently by @thewtex:

I feel an ethical obligation to structure this forum to promote open and reproducible science, which in this context entails open source.

It is also important to realize that open source alone is not enough. No software is perfect. Software may be buggy, poorly maintained, impenetrably complex, difficult to build, statistically unsound, produce irreproducible results… and still be open source. Furthermore: “Making something open isn’t a simple check box or button—it takes work, money, and time. Often those pushing for open access aren’t the ones who will have to implement it.” [source].

It is very challenging to write good software, and very challenging to pursue and communicate good science, even without any intentional secrets. We should do whatever we can to fight the replication crisis and enable science that is not only reproducible, but extensible.

As such, I personally do not support closed-source software projects being featured in the top-level navigation bar. But it would be great to see counterpoints argued here. It could also be possible to highlight the most useful and/or popular closed-source software projects in some other way—e.g. an expandable navigation drawer? Tag icons? Other ideas?


#48

I personally would also not want any commercial tools at the top-level. I just want to be sure that in case somebody asks X and somebody (from a company or not does not matter) answer and suggests to use commercial tool Y, it would be OK to do so.
And of course people can point out that Y is closed-source etc. and therefore should not be used or ask for clarification.

I am well aware of the issues with closed-source software but som lot of e people inside the imaging community vastly underestimate to real costs and IP-releated issues with developing software and how hard the competition is.

So most IP departments and controlling will only give you money for development if you can prove that competitors can not easily copy it…
Believe me, those discussion can be very tricky. I know what I am talking about.