Re: Heap stack allocation - Design and Architecture Discussion Boards

The Reactive Manifesto (https://www.reactivemanifesto.org/) is several paragraphs long and lists a large number of alleged characteristics and advantages of reactive programming; so many in fact, that reactive programming appears to be offered as some sort of panacea that will solve all of our problems forever.

To me it seems that most of the alleged advantages of reactive programming as listed in that manifesto are either false, or non-sequiturs, so despite what the manifesto says, I posit that the sole purpose of existence of reactive programming is performance, in other words saving threads.

Incidentally, this means that in any environment that offers virtual threads (for example, in Java starting from version 19) reactive programming is irrelevant.

Change my mind.

P.S.

Unfortunately, when people become salespersons for a certain cause, they seem to be never content with just mentioning the one game-changing advantage of their product over the competition; they seem to always want to throw as much as possible at the customer, hoping to make them buy; so, they tend to include a torrent of inconsequential or even entirely fictitious advantages, which often has the effect of drowning the one important advantage in the noise. For example I have seen this with microservices, whose lists of advantages are often twenty items long; most of them are preposterous, and almost all of them are nothing but filler, because in fact microservices only have one game-changing advantage, which is scalability, or two if we want to also count resilience.

Sorry, I'm not familiar with reactive systems but the reference you gave is just a bunch of gobbledygook. Platitudes and truisms not a design method make.

I spent many years in real-time data acquisition world and all the stuff mentioned in the "manifesto' was so obvious that no one would even bother to mention it:
- Responsive: show me a real time system that is not responsive
- Resilient: nah, we'll just let our system crash at the first bad input
- Elastic: we had to wait for 2014 for someone to bring up the subject because we never heard of the code 1201 and 1202 errors in '69[^]
- Message driven: another novelty... who would have thought?!

Laughable!

Mircea

It's Version 2 and published over eight years ago. The world has moved on ...

My apps reflect all the characteristics refered to as "reactive" though I never though of labeling them as such:

Responsive: sub 100ms response time
Resilient: graceful degradation in the face of system issues
Elastic: a frame rate that accomodates an increased load
Message driven: all operations that need to be async are so

So, "reactive" seems to be a new word to describe what should have been taken for granted in any worthwhile app.

"Before entering on an understanding, I have meditated for a long time, and have foreseen what might happen. It is not genius which reveals to me suddenly, secretly, what I have to say or to do in a circumstance unexpected by other people; it is reflection, it is meditation." - Napoleon I

Mike Nakis wrote:
reactive programming appears to be offered as some sort of panacea that will solve all of our problems forever.

Of course. Not much of an evangelical if one has a new idea and then says 'but this solves almost nothing so you should almost never use it. And especially not when considering long term maintenance costs.'

Mike Nakis wrote:
become salespersons for a certain cause, they seem to be never content with just mentioning the one game-changing advantage of their product over the competition; they seem to always want to throw as much as possible at the customer,

Also of course. As a salesperson ones income is derived from that. Probably not a good idea for the owner of a McDonald's to tell you that Burger King burgers are cheaper and taste better.

-------------------------------------------------
From the link

"Systems built as Reactive Systems are more flexible, loosely-coupled and scalable."<pre>

And with complexity that is significant.  And that is by far the most significant factor is any large scale enterprise.  No technology will fix that.

<pre>"The system responds in a timely manner if at all possible."<pre>

WTF?  What system of any sort does not do that?

<pre>"The system stays responsive in the face of failure."

I doubt that is a given. If the architecture, design and implementation does not specifically do that then there is nothing a specific technology can do to fix that.

"that ensures loose coupling,"<pre>

Which also insures complexity.

Componentizing a system, any system using any methodology, means that the individual component (pick your granularity) is easier to understand but it adds to the complexity of understanding and diagnosing problems that impact the enterprise.

I just wish all of the difficult problems that I have worked on for the last 20 years could have been easily solved by looking at one component.  The problems that were caused by a single component often could be diagnosed and fixed in less than a day.

<pre>"The largest systems in the world rely upon architectures based on these properties"

The largest systems in the world have been built over decades and have been based on many false starts and trial and error solutions to meet the specific business needs (and that is a very big plural) of the companies which hold those systems.

Small start ups cannot start with the assumption that they will need to support 1 billion users in 20 years because right now they are not even sure they have enough money to even run for 3 years. So they need something that works quickly no matter how it might fare in 20 years.

It's a raging tech sea. Many want anchors.

If you want a quick ride to the bottom just beeline to the nearest lighthouse looking thing like this while pretending it's a silver bullet. Sell it to management as such too.

It worked for agile! Even though that manifestos creators have basically since gone, "WTF, people? We weren't meaning to found a religion."

Not to say I think agile is even "bad". But snake oil, silver bullets, and fascist purists sure tend to be.

(I brought this up as a sidetrack in a Lounge thread. Even though I ask out of curiosity only, not any 'need' or immediate problem, the question is on the technical side, so I move it over here Smile | :)

There is no real reason why stack frames are allocated edge to edge! The stack frame head 'previous frame' pointer could go anywhere. Stack relative addressing never goes outside the current frame. Traditionally, with a million objects, each running its own thread, each of them has allotted a stack for its maximum call depth. This can bind up quite a few megabytes that are hardly if ever used, most certainly not all at the same time. Never ever are every single thread preempted at its very deepest nesting of calls at exactly the same time. In a non-preemptive, co-routine based system, they won't all yield (or whatever it is called in your favorite co-routine package) at the maximum stack depth, all at the same time.

Stack frames could be allocated from the heap, with space is occupied only as long as a function call is active. Then, a given amount of RAM would be capable of handling a much larger number of threads. Especially in event driven architectures, a large fraction of threads idle at a low stack level. They receive an event and go into a nest of calls for handling it, and then return to the low stack level waiting for the next event. The thread might do some sort of yield halfway, but the great majority of threads would be back to 'ground base', or at a moderate call nesting level, most of the time.

The argument against heap allocation is of course the processing cost. There are (/were) machines offering micro coded entry point instructions doing stack allocation from a buddy heap, essentially unlinking the top element from the free list for the appropriate frame size. Function return links the frame back to the top of the list. This requires a couple of memory cycles for each operation. What is that on the total instruction budget for a modern style method call?

Even though you with desktop PCs can just plug in another 64 GiB of RAM to satisfy stack needs, in other systems (such as embedded), this is not always possible. Yet, lots of programmers recoil in horror over buddy allocation: It causes lots internal fragmentation! 25% with binary buddies! Well, but how much can you save in stack requirements? Besides, lots of architectures demand word, doubleword or paragraph stack alignment anyway - that leads to internal fragmentation/waste as well!

If the allocation cost worries you, lots of optimizations are possible, in particular if frame allocation is the responsibility of the caller. E.g. if a function's stack frame has a lot of unused (internal fragmentation) space calls a function asking for a small frame, it could fit in that internal fragmentation space, not requiring another heap allocation. There are dozens of such optimization.

I never heard of any software (compiler, run time system) allocating stack frames on the heap. I never heard of anyone using the micro coded buddy stack frame allocation on the one architecture I know of providing it. Is that because I am ignorant? Smile | Smile | :)

Is heap stack allocation 'commonplace' anywhere? Or has it been considered and rejected? If that is the case, what were the arguments for rejecting it?

Lot of text there with no context.

trønderen wrote:
There is no real reason why stack frames are allocated edge to edge!

In C/C++ there were library calls that allowed one to chop the stack off. To reset it to a specific point. I never used it, so I don't know why it existed. Might have been performance or error handling. That call would have certainly been easier (and faster) if frames were next to each other.

However I also believe that is not the case for Java. I believe I looked at the source code for that at one point and it was just allocating on the heap.

trønderen wrote:
Stack relative addressing never goes outside the current frame

False.

In C and C++ one can overwrite the frame. I have in fact done so in the past.
Since both Java and C# support native code overwriting the frame is also possible there.

trønderen wrote:
Stack frames could be allocated from the heap, with space is occupied only as long as a function call is active

That is somewhat simplistic. Allocations of any sort still require a mechanism to deallocate it.

So for example a standard in C/C++ (at least when I looked at source code) was that local variable were allocated. Might have even been in the stack frame. Then the deallocate was when the frame was freed on exit.
That is why they were (or are) on the stack frame. Because the deallocation is already there.

Heap deallocation on the other hand is a different mechanism.

trønderen wrote:
Then, a given amount of RAM would be capable of handling a much larger number of threads.

As stated that is wrong. Memory is not the only limitation for the number of threads. Not to mention that no developer should ever consider the idea that an unlimited or large number of threads is a 'good' idea.

trønderen wrote:
Even though you with desktop PCs can just plug in another 64 GiB of RAM to satisfy stack needs

Just noting that I have never seen a computer that actually allows for the full addressable range of 64 bits. Certainly no desktop PC does that. So one cannot just keep plugging it in.

And cloud servers are always limited also. Sometimes to very surprising low limits.

trønderen wrote:
(such as embedded), this is not always possible. Yet, lots of programmers recoil in horror

Because all resources, not just memory, is scarce. If you add complication to the processing model then it will require more resources. There is no way around that.

trønderen wrote:
If the allocation cost worries you, lots of optimizations are possible,

The discussion however is not about me (and the general audience) but rather you.

If you want to fix it then do so.

I have in the past replaced the entire heap in C compiler eco system.
Certainly you could also do the same for the Stack Frame allocation. So do it. I suspect you will find that the deallocation part is going be complicated but doable.

trønderen wrote:
I never heard of any software (compiler, run time system) allocating stack frames on the heap

As I suggested when I looked I found Java (JVM source code) to do that.

Following has a response that suggests the same thing.

https://stackoverflow.com/questions/26741925/is-frame-in-jvm-heap-allocated-or-stack-allocated[^]

jschell wrote:
Lot of text there with no context.

What kind of 'context' are you requesting? A call stack is a general concept, employed in practically speaking all current programming languages. Do you want the context limited to one specific language? One specific hardware platform?

In C/C++ there were library calls that allowed one to chop the stack off. To reset it to a specific point. I never used it, so I don't know why it existed.

I never knew of anything like that, and your reference is so unspecific (no function names etc.) that it is difficult to do a search. From what you write, it may sound like a mechanism intended for threads requiring a lot of stack space during initialization, and then enter a 'working' state with small stack requirements, so the stack limit is reduced after init, to make the address space above it available to other threads (started at a later time). I am guessing now, and may be completely wrong, but I can't think of many other uses that makes sense (based on the information you provide).

trønderen wrote: Stack relative addressing never goes outside the current frame
False. In C and C++ one can overwrite the frame.

I am not sure what you mean by 'overwriting', suspecting that you refer to something similar to out-of-bounds array indexing: If you put a somearray[1] at the very top of the stack, you may index it beyond the limit of the current frame. That is as bad as any other out-of-bounds indexing! Conceptually, you are still within the same stack frame, you are just pretending that the frame is larger than initially allocated. There is no guarantee that the space you are trying to (mis)use is available; you might hit the stack limit.

In languages with a static link (I haven't been working with any more recent one than Pascal; most newer languages don't have a static link), code of an inner function may address locations in frames lower on the stack. However, they do not do this by negative offsets (or positive, if the stack is growing downwards) from their own frame pointer. Rather, they use the static link to find the stack frame of the outer function, and do relative addressing from that frame base address - as if it was a pointer to a struct. The addressing never goes outside that frame.

Allocations of any sort still require a mechanism to deallocate it. [...] That is why they were (or are) on the stack frame. Because the deallocation is already there.

Eh? What's the problem? When a method completes, it returns, whether the frame was allocated edge-by-edge or on the heap. In the former case, you adjust the top of stack register to the top of the previous frame. In the heap case, you move the freelist[x] value to the 'next' field (in the freelist interpretation of 'next') of the frame and store the address of the frame in freelist[x]. In both cases, you set the current stack frame pointer to the previous frame. I don't see any big difference at all!

trønderen wrote: Then, a given amount of RAM would be capable of handling a much larger number of threads.
As stated that is wrong. Memory is not the only limitation for the number of threads.

You certainly have not explained why this would be 'wrong'. There may certainly be other limitations on the number of threads, but memory is a significant one. Not in all cases, but quite often. Other limitations may be e.g. number/size of thread descriptors; this is highly OS / runtime system dependent, and need not be neither a hard nor a significant limit.

Not to mention that no developer should ever consider the idea that an unlimited or large number of threads is a 'good' idea.

It may be a bad idea, e.g. if the OS/RTS puts a hard limit on the number of descriptors, or the memory cost for each thread is large. Heap allocation may reduce the average memory cost significantly, reducing one limiting factor.

Just noting that I have never seen a computer that actually allows for the full addressable range of 64 bits. Certainly no desktop PC does that.

You are certainly confusing 64 address bits with 64 gigabyte, which can be addressed with 36 bits. Current mainboards can handle up to 256 Gi.

And cloud servers are always limited also. Sometimes to very surprising low limits.

You are repeating the point I made, except that I referred to embedded systems, not to cloud servers. I used '64 Gi' not as a specific value, just to emphasize that option to add more memory can be extremely limited, compared to a plain desktop PC - most tower cabinet PC users still have empty memory slots on the mainboard and can add more RAM. Not so in an IoT chip.

Because all resources, not just memory, is scarce. If you add complication to the processing model then it will require more resources. There is no way around that.

You certainly haven't justified why unlinking a block from a freelist rather than adding to the top of stack register would 'add complication to the processing model'. It is so simple that at least one machine provided it as a single entry point instruction (and a corresponding return instruction), everything else was 100% identical. The purpose of it is to save (memory) resources.

I suspect you will find that the deallocation part is going be complicated but doable.

In the ND-500 architecture, it was a single instruction covering both deallocation and return, copying the current freelist[x] into the freed frame and the frame address into freelist[x] - that is not 'complicated but doable'! (The instruction also did all the other return stuff.)

For all heap systems, you must handle splitting of large free blocks to get one of the size you want, and when space runs out, you must recombine smaller, adjacent free blocks. In buddy systems, both are fairly trivial operations. Furthermore, recombination can be done incrementally, e.g when waiting for I/O and no thread is runnable. It is also cheap, especially binary buddy, both for the splitting (as implemented in the ND-500) and recombination.

Following has a response that suggests the same thing.
https://stackoverflow.com/questions/26741925/is-frame-in-jvm-heap-allocated-or-stack-allocated[^]

I find it slightly funny that you spend a lot of space and energy to tell me how non-workable my proposal for a heap allocated stack is, and then round it off with a reference to a post documenting that it certainly is a viable alternative and has been used for language as widespread as Java Smile | :)

trønderen wrote:
From what you write, it may sound like a mechanism intended for threads requiring a lot of stack space during initialization,

Incorrect. C existed before threads.

The call stack existed then only for controlling the method calling process.

Even now threads on C are more of a add on. So threads must work within the way C does stuff.

trønderen wrote:
suspecting that you refer to something similar to out-of-bounds array indexing:

No I wrote what I meant. You are referring to reading. I was talking about writing.

trønderen wrote:
There may certainly be other limitations on the number of threads, but memory is a significant one.

For every resource available to a running program memory is the one of which there is the 'most' available. Every other resource of any kind is going to be much much smaller.

As a specific example of that it doesn't matter if a language provides a way to provide for an unlimited number of threads when the OS will not. And never will because nothing in computing is unlimited.

trønderen wrote:
You are certainly confusing 64 address bits with 64 gigabyte,

No confusion on my part. You suggested that someone could just plug in an additional 64 gig of memory. That would of course only be to a 64 bit (or higher) CPU. However, as I said, there are very real practical limits on existing computers (the physical boxes and cloud ones) which mean that your suggestion is not generally true.

trønderen wrote:
I find it slightly funny that you spend a lot of space and energy to tell me how non-workable my proposal for a heap allocated stack is

I addressed specific points that you made, because as I pointed out in the very first sentence of my response, you long post seemed to ramble on quite a bit.

So for example in the first part of my response you claimed there was never any reason for doing it - so I showed that there was. I based that on my understanding the history of programming and more specifically on the history of the C programming language.

But in general you claimed that no one does heap based stack allocations but I pointed out that they do. So certainly nothing contradictory in my statements in posting a link that says that.

trønderen wrote:
In languages with a static link (I haven't been working with any more recent one than Pascal; most newer languages don't have a static link), code of an inner function may address locations in frames lower on the stack. However, they do not do this by negative offsets (or positive, if the stack is growing downwards) from their own frame pointer. Rather, they use the static link to find the stack frame of the outer function, and do relative addressing from that frame base address - as if it was a pointer to a struct. The addressing never goes outside that frame.

In the late 1970s, I had a project to optimise expression evaluation in Pascal (based on the Pascal PCode compiler). As you surmised, in the PCode compiler, accessing higher nested values from lower nested functions did skip down the stack frame back pointers (one skip per depth of nesting); but that is an implementation technique, it is not part of the language definition. I guessed that it could be improved by maintaining a vector of nested stack frame start locations to access stack specific offsets directly using this vector. This proved to be a useful optimisation for evaluating expressions. However, when I ran comparisons on a large Pascal program (my test large program was the original version of the PCode compiler) I discovered that whilst the overheads for evaluating expressions were reduced, the number of times that the test program actually accessed variables in higher scopes was so low that the overhead of maintaining the vector outweighed the advantages of having the vector.

Branching stacks aka spaghetti stacks are used in the implementations of some functional languages. They work basically how you suggest they could work, plus the added detail of allowing "sibling frames" to exist simultaneously.

Go supposedly used to use segmented/split stacks (which also work roughly as you suggested, not on a per-frame basis exactly but more as a list of blocks), but later apparently switched to just copying the whole damn thing to keep it contiguous (which still leaves "the stack(TM)" as essentially a bunch of arraylists that exist on the heap). Other userspace-threading solutions may use similar techniques for their stacks for "fake threads", but I don't really pay attention to what happens in that space, hell I don't really know what Go is doing either, I just remembered that it did something funny and spent 5 minutes on google to look it up.

By the way, the way stacks normally use, is by reserving a bunch of contiguous virtual address space, and lazily commit pages of actual physical memory only as needed. Which is why large stack allocations involve a call to _chkstk, to ensure that every page is touched in order, to avoid missing the guard page. So already with the stack working the way it normally does, you're probably not paying the full memory cost for each stack. But there is no mechanism to shrink the stack. Once a page has been touched once, you keep it.

If you do dynamic allocation of physical pages to a stack, each stack has a minimum memory cost of one RAM page (typically 4 Kibytes), and a similar increment. If the typical thread nests deeply, and you have a few threads, this is OK, but if you rather go for tiny, lightweight threads and lots of them, 4 Ki might lead to a lot of internal fragmentation.

One disadvantage: If you start new threads very frequently, and each start operation leads to a stack page fault, causing allocation of a new physical page and updating of MMS tables, then the cost of starting new threads, as well as the delay before the new thread is running, will increase noticeable, to phrase it politely.

Btw: I have seen such 'on demand' allocation of physical memory pages not just for stacks, but also for heap memory.

Small association: A long time ago, CPU architects were arguing whether stacks ought to grow upwards or downwards. One of the downwards arguments was that if you put the stack at the bottom of the address range, a stack overflow could be detected by an attempt to address below address zero, by the sign bit. If that logic was still in use, it would prevent an 'on demand extension' of the stack.

trønderen wrote:
One disadvantage: If you start new threads very frequently, and each start operation leads to a stack page fault, causing allocation of a new physical page and updating of MMS tables, then the cost of starting new threads, as well as the delay before the new thread is running, will increase noticeable, to phrase it politely.

That however is a architecture/design problem. Not a technical one.

Computers have limits and always will limits. So one can never create an architecture/design that is based on unlimited resources. In fact one should plan on smaller than actual resources to give one room for unexpected problems.

I have been working on a WPF app for a client for 3 years. They have someone on site who verifies changes I make.

What I've been doing up to now is making a copy of the SQL datbase on their server, restoring to a test DB, and appling any SQL changes to it. I then copy all test files to their system into a test folder. The have a Test Mode icon on the desktop which allows them to run the test environment.

Once all changes are verified I repeate the above for their Production environment.

This has worked well, but I'd like to have a way to test different changes independently. A test environment for each branch would be nice, but that means duplicating what I wrote above for each new branch. Something really easy to set up and take down as branches change would be great.

I'm open to ideas.

If it's not broken, fix it until it is.
Everything makes sense in someone's mind.
Ya can't fix stupid.

modified 28-Dec-22 18:11pm.

The above is not all that clear. Perhaps because the process has not actually been defined.

Kevin Marois wrote:
What I've been doing up to now is

This is a 'deployment' or a 'release'

Kevin Marois wrote:
A test environment for each branch would be nice,

Ok but what does that have to do with a 'deployment'?

Generally a branch is used to work on feature/fix/etc. So lets say right now there is a new feature and two bug fixes that you need to work on.

I am going to presume you are using source control. So locally you can work on each as a branch. When a branch is complete then you merge it to the main branch in source control. You only 'deliver' the main branch. So if you have finished the feature and one bug when you decide to deliver than you using the main branch to do that. The other bug is in the branch that is not complete and not merged.

Items in the above
1. Depending on how you deliver it you might need to rebuild the main branch. This is a process step that you would need to do. Probably should have an actual list of steps. Important step is that you should always extract from main branch even if you think it is already there.
2. SQL, or however you update the database, should be checked into source control also.

Kevin Marois wrote:
now is making a copy of the SQL datbase on their server,...

Steps for testing
1. Create test database. Specific steps are irrelevant.
2. Apply NEW updates the test database.
3. Deploy NEW application to test folder. Specific steps are irrelevant.
4. The icon on desktop to test

Steps for production
1. Apply NEW updates the real database.
2. Deploy NEW application to real folder. Specific steps are irrelevant.

In the above the two steps match the appropriate previous steps except that it applies to a different destination. So you can certainly automate that using a script (linux or windows.) You can also create an application (installer) to do the same thing.

Don't forget potential errors. For example someone accidently marks a file in the real folder is read only and so your script/installer can't copy the new file in. Might just be a matter of manually inspecting the output but there are other ways.

Flyway... I'd use Flyway.

Whether using repeatable migrations or versioned + undo you could accomplish what I'll call "feature-branched DB state management" which is what it sounds like you want.

I know commanding and converters have been a staple of MVVM pattern and XAML interfaces, but it seems like WinUI 3 is moving away from those, particularly in terms of the functionality offered (direct binding to functions and events) and seemingly encouragement in the documentation to use them instead of commands and converters.

I'm in the middle of refactoring a WPF application to WinUI 3, and after writing several XAML pages, it looks like it's going to be possible to avoid commands and converters altogether. So is it an MVVM sin to wholly replace commands and converters with x:Bind functions?

Randy

MS samples are meant to show language features. They don't "do MVVM" unless the article is about MVVM. x:Bind provides some additional type checking at design time and some performance advantages versus "Binding"; with trade-offs. Nothing to do with MVVM in particular.

"Before entering on an understanding, I have meditated for a long time, and have foreseen what might happen. It is not genius which reveals to me suddenly, secretly, what I have to say or to do in a circumstance unexpected by other people; it is reflection, it is meditation." - Napoleon I

Thanks, Gerry. My WPF application code was fairly consistent in its application of Binding with commands and converters, and I'd like to continue that level of consistency as I refactor it to WinUI 3.

I'd like to take advantage of the x:Bind performance improvements, but also remain consistent in the way it's used. In other words, I'd prefer to continue to wholly use Binding (to DataContext) with traditional commands and converters, wholly use x:Bind (to code-behind objects) with traditional commands and converter classes, or wholly use x:Bind with functions (no commands or converters).

This may not make sense or be possible, I just haven't gotten far enough into it to tell. So far I haven't found any showstoppers, and that's why I'm curious. But I'm also only about 10% into it.

I'm working on a WPF app that will connect to Google and retrieve contacts. Right now, the Google key info is all hardcoded unencrypted into the app and is a security risk. I'd like to refactor this so that the key is not compiled in.

One idea we had was to make a backend call to our server to retrieve the key, then use it to connect to Google.

What's the right way to do this?

If it's not broken, fix it until it is.
Everything makes sense in someone's mind.
Ya can't fix stupid.

modified 13-Dec-22 14:24pm.

The server connects to / queries Google; the client makes the Google request via the server.

"Before entering on an understanding, I have meditated for a long time, and have foreseen what might happen. It is not genius which reveals to me suddenly, secretly, what I have to say or to do in a circumstance unexpected by other people; it is reflection, it is meditation." - Napoleon I

OK.

THe issue is that the Security Key & Secret are stored local in the client. They have to be passed. I'm asking about how to design this for security

If it's not broken, fix it until it is.
Everything makes sense in someone's mind.
Ya can't fix stupid.

But if the key is shared, shouldn't it be on the server? You route the queries through the server; the query runs from the server; the client never needs to see the key; only the results.

"Before entering on an understanding, I have meditated for a long time, and have foreseen what might happen. It is not genius which reveals to me suddenly, secretly, what I have to say or to do in a circumstance unexpected by other people; it is reflection, it is meditation." - Napoleon I

You pass a Key and Secret to the server. We're trying to avoid this

If it's not broken, fix it until it is.
Everything makes sense in someone's mind.
Ya can't fix stupid.